AIEnhancedStrategy
This commit is contained in:
zhangkun9038@dingtalk.com
parent
ef3b1045b2
commit
9a87fad909
64
config_examples/aienhance_config.json
Normal file
64
config_examples/aienhance_config.json
Normal file
@ -0,0 +1,64 @@
|
||||
{
|
||||
"max_open_trades": 3,
|
||||
"stake_currency": "USDT",
|
||||
"stake_amount": 100,
|
||||
"tradable_balance_ratio": 0.99,
|
||||
"fiat_display_currency": "USD",
|
||||
"timeframe": "5m",
|
||||
"exchange": {
|
||||
"name": "okx",
|
||||
"key": "your_api_key",
|
||||
"secret": "your_api_secret",
|
||||
"password": "your_api_password",
|
||||
"type": "spot",
|
||||
"ccxt_config": {
|
||||
"enableRateLimit": true,
|
||||
"options": {
|
||||
"defaultType": "spot"
|
||||
}
|
||||
},
|
||||
"ccxt_async_config": {
|
||||
"enableRateLimit": true,
|
||||
"timeout": 30000
|
||||
}
|
||||
},
|
||||
"pairlists": [
|
||||
{
|
||||
"method": "StaticPairList",
|
||||
"allow_inactive": false,
|
||||
"number_assets": 5,
|
||||
"pairs": ["BTC/USDT", "ETH/USDT", "BNB/USDT", "XRP/USDT", "SOL/USDT"]
|
||||
}
|
||||
],
|
||||
"strategy": "AIEnhancedStrategy",
|
||||
"freqai": {
|
||||
"enabled": true,
|
||||
"live_retrain_hours": 1,
|
||||
"data_kitchen_path": "freqai/data_kitchen/",
|
||||
"purge_old_data_seconds": 86400,
|
||||
"feature_parameters": {
|
||||
"include_timeframes": ["5m", "15m", "1h"],
|
||||
"include_shifted_candles": 2,
|
||||
"DI_threshold": 0.85,
|
||||
"label_period_candles": 20,
|
||||
"include_volume_data": true,
|
||||
"include_bidirectional_training_data": true
|
||||
},
|
||||
"model_training_parameters": {
|
||||
"model_type": "LightGBMRegressor",
|
||||
"n_estimators": 100,
|
||||
"learning_rate": 0.1,
|
||||
"max_depth": 10,
|
||||
"early_stopping_rounds": 10
|
||||
}
|
||||
},
|
||||
"dry_run": true,
|
||||
"dry_run_wallet": 1000,
|
||||
"cancel_open_orders_on_exit": false,
|
||||
"protections": [],
|
||||
"internals": {
|
||||
"process_only_new_candles": true
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -67,8 +67,8 @@ services:
|
||||
backtesting
|
||||
--logfile /freqtrade/user_data/logs/freqtrade.log
|
||||
--freqaimodel XGBoostRegressor
|
||||
--config /freqtrade/config_examples/ml_config.json
|
||||
--config /freqtrade/config_examples/aienhance_config.json
|
||||
--strategy-path /freqtrade/templates
|
||||
--strategy MLBasedSentimentStrategy
|
||||
--strategy AIEnhancedStrategy
|
||||
--timerange 20250401-20250420
|
||||
--cache none
|
||||
|
||||
59
freqtrade/templates/AIEnhancedStrategy.py
Normal file
59
freqtrade/templates/AIEnhancedStrategy.py
Normal file
@ -0,0 +1,59 @@
|
||||
from freqtrade.strategy.interface import IStrategy
|
||||
from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
|
||||
from pandas import DataFrame
|
||||
import numpy as np
|
||||
import logging
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
class AIEnhancedStrategy(IStrategy):
|
||||
INTERFACE_VERSION = 3
|
||||
|
||||
can_short = False # 只做多
|
||||
timeframe = '5m'
|
||||
|
||||
process_only_new_candles = True
|
||||
use_exit_signal = True
|
||||
exit_profit_only = False
|
||||
ignore_roi_if_entry_signal = False
|
||||
|
||||
def feature_engineering_expand_all(self, dataframe: DataFrame, period: int,
|
||||
metadata: dict, **kwargs) -> DataFrame:
|
||||
# 自定义特征工程
|
||||
for col in ['open', 'high', 'low', 'close', 'volume']:
|
||||
dataframe[f'{col}_pct_change'] = dataframe[col].pct_change(periods=period)
|
||||
return dataframe
|
||||
|
||||
def set_freqai_targets(self, dataframe: DataFrame, metadata: dict, **kwargs) -> DataFrame:
|
||||
# 目标变量：未来n根K线的收盘价变化
|
||||
dataframe['&s-close_pct'] = dataframe['close'].pct_change(periods=5).shift(-5)
|
||||
return dataframe
|
||||
|
||||
def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
|
||||
dk = self.freqai_info.get("dk", None)
|
||||
if not dk:
|
||||
raise ValueError("FreqaiDataKitchen is not available.")
|
||||
|
||||
dataframe = dk.feature_engineering_standard(dataframe, metadata, self)
|
||||
|
||||
if self.config["runmode"].value in ("live", "dry_run"):
|
||||
dataframe = dk.live_models(self, dataframe, metadata=metadata)
|
||||
|
||||
return dataframe
|
||||
|
||||
def populate_entry_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
|
||||
# 预测值大于阈值时开多仓
|
||||
dataframe.loc[
|
||||
(dataframe["&s-close_pct"] > 0.002),
|
||||
"enter_long"
|
||||
] = 1
|
||||
return dataframe
|
||||
|
||||
def populate_exit_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
|
||||
# 预测值小于负阈值时平仓
|
||||
dataframe.loc[
|
||||
(dataframe["&s-close_pct"] < -0.001),
|
||||
"exit_long"
|
||||
] = 1
|
||||
return dataframe
|
||||
|
||||
423
md/freqai-configuration.md
Normal file
423
md/freqai-configuration.md
Normal file
@ -0,0 +1,423 @@
|
||||
# Configuration
|
||||
|
||||
FreqAI is configured through the typical [Freqtrade config file](configuration.md) and the standard [Freqtrade strategy](strategy-customization.md). Examples of FreqAI config and strategy files can be found in `config_examples/config_freqai.example.json` and `freqtrade/templates/FreqaiExampleStrategy.py`, respectively.
|
||||
|
||||
## Setting up the configuration file
|
||||
|
||||
Although there are plenty of additional parameters to choose from, as highlighted in the [parameter table](freqai-parameter-table.md#parameter-table), a FreqAI config must at minimum include the following parameters (the parameter values are only examples):
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"enabled": true,
|
||||
"purge_old_models": 2,
|
||||
"train_period_days": 30,
|
||||
"backtest_period_days": 7,
|
||||
"identifier" : "unique-id",
|
||||
"feature_parameters" : {
|
||||
"include_timeframes": ["5m","15m","4h"],
|
||||
"include_corr_pairlist": [
|
||||
"ETH/USD",
|
||||
"LINK/USD",
|
||||
"BNB/USD"
|
||||
],
|
||||
"label_period_candles": 24,
|
||||
"include_shifted_candles": 2,
|
||||
"indicator_periods_candles": [10, 20]
|
||||
},
|
||||
"data_split_parameters" : {
|
||||
"test_size": 0.25
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
A full example config is available in `config_examples/config_freqai.example.json`.
|
||||
|
||||
!!! Note
|
||||
The `identifier` is commonly overlooked by newcomers, however, this value plays an important role in your configuration. This value is a unique ID that you choose to describe one of your runs. Keeping it the same allows you to maintain crash resilience as well as faster backtesting. As soon as you want to try a new run (new features, new model, etc.), you should change this value (or delete the `user_data/models/unique-id` folder. More details available in the [parameter table](freqai-parameter-table.md#feature-parameters).
|
||||
|
||||
## Building a FreqAI strategy
|
||||
|
||||
The FreqAI strategy requires including the following lines of code in the standard [Freqtrade strategy](strategy-customization.md):
|
||||
|
||||
```python
|
||||
# user should define the maximum startup candle count (the largest number of candles
|
||||
# passed to any single indicator)
|
||||
startup_candle_count: int = 20
|
||||
|
||||
def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
|
||||
|
||||
# the model will return all labels created by user in `set_freqai_targets()`
|
||||
# (& appended targets), an indication of whether or not the prediction should be accepted,
|
||||
# the target mean/std values for each of the labels created by user in
|
||||
# `set_freqai_targets()` for each training period.
|
||||
|
||||
dataframe = self.freqai.start(dataframe, metadata, self)
|
||||
|
||||
return dataframe
|
||||
|
||||
def feature_engineering_expand_all(self, dataframe: DataFrame, period, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This function will automatically expand the defined features on the config defined
|
||||
`indicator_periods_candles`, `include_timeframes`, `include_shifted_candles`, and
|
||||
`include_corr_pairs`. In other words, a single feature defined in this function
|
||||
will automatically expand to a total of
|
||||
`indicator_periods_candles` * `include_timeframes` * `include_shifted_candles` *
|
||||
`include_corr_pairs` numbers of features added to the model.
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
:param period: period of the indicator - usage example:
|
||||
dataframe["%-ema-period"] = ta.EMA(dataframe, timeperiod=period)
|
||||
"""
|
||||
|
||||
dataframe["%-rsi-period"] = ta.RSI(dataframe, timeperiod=period)
|
||||
dataframe["%-mfi-period"] = ta.MFI(dataframe, timeperiod=period)
|
||||
dataframe["%-adx-period"] = ta.ADX(dataframe, timeperiod=period)
|
||||
dataframe["%-sma-period"] = ta.SMA(dataframe, timeperiod=period)
|
||||
dataframe["%-ema-period"] = ta.EMA(dataframe, timeperiod=period)
|
||||
|
||||
return dataframe
|
||||
|
||||
def feature_engineering_expand_basic(self, dataframe: DataFrame, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This function will automatically expand the defined features on the config defined
|
||||
`include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`.
|
||||
In other words, a single feature defined in this function
|
||||
will automatically expand to a total of
|
||||
`include_timeframes` * `include_shifted_candles` * `include_corr_pairs`
|
||||
numbers of features added to the model.
|
||||
|
||||
Features defined here will *not* be automatically duplicated on user defined
|
||||
`indicator_periods_candles`
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
dataframe["%-pct-change"] = dataframe["close"].pct_change()
|
||||
dataframe["%-ema-200"] = ta.EMA(dataframe, timeperiod=200)
|
||||
"""
|
||||
dataframe["%-pct-change"] = dataframe["close"].pct_change()
|
||||
dataframe["%-raw_volume"] = dataframe["volume"]
|
||||
dataframe["%-raw_price"] = dataframe["close"]
|
||||
return dataframe
|
||||
|
||||
def feature_engineering_standard(self, dataframe: DataFrame, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This optional function will be called once with the dataframe of the base timeframe.
|
||||
This is the final function to be called, which means that the dataframe entering this
|
||||
function will contain all the features and columns created by all other
|
||||
freqai_feature_engineering_* functions.
|
||||
|
||||
This function is a good place to do custom exotic feature extractions (e.g. tsfresh).
|
||||
This function is a good place for any feature that should not be auto-expanded upon
|
||||
(e.g. day of the week).
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
usage example: dataframe["%-day_of_week"] = (dataframe["date"].dt.dayofweek + 1) / 7
|
||||
"""
|
||||
dataframe["%-day_of_week"] = (dataframe["date"].dt.dayofweek + 1) / 7
|
||||
dataframe["%-hour_of_day"] = (dataframe["date"].dt.hour + 1) / 25
|
||||
return dataframe
|
||||
|
||||
def set_freqai_targets(self, dataframe: DataFrame, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
Required function to set the targets for the model.
|
||||
All targets must be prepended with `&` to be recognized by the FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the targets
|
||||
usage example: dataframe["&-target"] = dataframe["close"].shift(-1) / dataframe["close"]
|
||||
"""
|
||||
dataframe["&-s_close"] = (
|
||||
dataframe["close"]
|
||||
.shift(-self.freqai_info["feature_parameters"]["label_period_candles"])
|
||||
.rolling(self.freqai_info["feature_parameters"]["label_period_candles"])
|
||||
.mean()
|
||||
/ dataframe["close"]
|
||||
- 1
|
||||
)
|
||||
return dataframe
|
||||
```
|
||||
|
||||
Notice how the `feature_engineering_*()` is where [features](freqai-feature-engineering.md#feature-engineering) are added. Meanwhile `set_freqai_targets()` adds the labels/targets. A full example strategy is available in `templates/FreqaiExampleStrategy.py`.
|
||||
|
||||
!!! Note
|
||||
The `self.freqai.start()` function cannot be called outside the `populate_indicators()`.
|
||||
|
||||
!!! Note
|
||||
Features **must** be defined in `feature_engineering_*()`. Defining FreqAI features in `populate_indicators()`
|
||||
will cause the algorithm to fail in live/dry mode. In order to add generalized features that are not associated with a specific pair or timeframe, you should use `feature_engineering_standard()`
|
||||
(as exemplified in `freqtrade/templates/FreqaiExampleStrategy.py`).
|
||||
|
||||
## Important dataframe key patterns
|
||||
|
||||
Below are the values you can expect to include/use inside a typical strategy dataframe (`df[]`):
|
||||
|
||||
| DataFrame Key | Description |
|
||||
|------------|-------------|
|
||||
| `df['&*']` | Any dataframe column prepended with `&` in `set_freqai_targets()` is treated as a training target (label) inside FreqAI (typically following the naming convention `&-s*`). For example, to predict the close price 40 candles into the future, you would set `df['&-s_close'] = df['close'].shift(-self.freqai_info["feature_parameters"]["label_period_candles"])` with `"label_period_candles": 40` in the config. FreqAI makes the predictions and gives them back under the same key (`df['&-s_close']`) to be used in `populate_entry/exit_trend()`. <br> **Datatype:** Depends on the output of the model.
|
||||
| `df['&*_std/mean']` | Standard deviation and mean values of the defined labels during training (or live tracking with `fit_live_predictions_candles`). Commonly used to understand the rarity of a prediction (use the z-score as shown in `templates/FreqaiExampleStrategy.py` and explained [here](#creating-a-dynamic-target-threshold) to evaluate how often a particular prediction was observed during training or historically with `fit_live_predictions_candles`). <br> **Datatype:** Float.
|
||||
| `df['do_predict']` | Indication of an outlier data point. The return value is integer between -2 and 2, which lets you know if the prediction is trustworthy or not. `do_predict==1` means that the prediction is trustworthy. If the Dissimilarity Index (DI, see details [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di)) of the input data point is above the threshold defined in the config, FreqAI will subtract 1 from `do_predict`, resulting in `do_predict==0`. If `use_SVM_to_remove_outliers` is active, the Support Vector Machine (SVM, see details [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm)) may also detect outliers in training and prediction data. In this case, the SVM will also subtract 1 from `do_predict`. If the input data point was considered an outlier by the SVM but not by the DI, or vice versa, the result will be `do_predict==0`. If both the DI and the SVM considers the input data point to be an outlier, the result will be `do_predict==-1`. As with the SVM, if `use_DBSCAN_to_remove_outliers` is active, DBSCAN (see details [here](freqai-feature-engineering.md#identifying-outliers-with-dbscan)) may also detect outliers and subtract 1 from `do_predict`. Hence, if both the SVM and DBSCAN are active and identify a datapoint that was above the DI threshold as an outlier, the result will be `do_predict==-2`. A particular case is when `do_predict == 2`, which means that the model has expired due to exceeding `expired_hours`. <br> **Datatype:** Integer between -2 and 2.
|
||||
| `df['DI_values']` | Dissimilarity Index (DI) values are proxies for the level of confidence FreqAI has in the prediction. A lower DI means the prediction is close to the training data, i.e., higher prediction confidence. See details about the DI [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di). <br> **Datatype:** Float.
|
||||
| `df['%*']` | Any dataframe column prepended with `%` in `feature_engineering_*()` is treated as a training feature. For example, you can include the RSI in the training feature set (similar to in `templates/FreqaiExampleStrategy.py`) by setting `df['%-rsi']`. See more details on how this is done [here](freqai-feature-engineering.md). <br> **Note:** Since the number of features prepended with `%` can multiply very quickly (10s of thousands of features are easily engineered using the multiplictative functionality of, e.g., `include_shifted_candles` and `include_timeframes` as described in the [parameter table](freqai-parameter-table.md)), these features are removed from the dataframe that is returned from FreqAI to the strategy. To keep a particular type of feature for plotting purposes, you would prepend it with `%%` (see details below). <br> **Datatype:** Depends on the feature created by the user.
|
||||
| `df['%%*']` | Any dataframe column prepended with `%%` in `feature_engineering_*()` is treated as a training feature, just the same as the above `%` prepend. However, in this case, the features are returned back to the strategy for FreqUI/plot-dataframe plotting and monitoring in Dry/Live/Backtesting <br> **Datatype:** Depends on the feature created by the user. Please note that features created in `feature_engineering_expand()` will have automatic FreqAI naming schemas depending on the expansions that you configured (i.e. `include_timeframes`, `include_corr_pairlist`, `indicators_periods_candles`, `include_shifted_candles`). So if you want to plot `%%-rsi` from `feature_engineering_expand_all()`, the final naming scheme for your plotting config would be: `%%-rsi-period_10_ETH/USDT:USDT_1h` for the `rsi` feature with `period=10`, `timeframe=1h`, and `pair=ETH/USDT:USDT` (the `:USDT` is added if you are using futures pairs). It is useful to simply add `print(dataframe.columns)` in your `populate_indicators()` after `self.freqai.start()` to see the full list of available features that are returned to the strategy for plotting purposes.
|
||||
|
||||
## Setting the `startup_candle_count`
|
||||
|
||||
The `startup_candle_count` in the FreqAI strategy needs to be set up in the same way as in the standard Freqtrade strategy (see details [here](strategy-customization.md#strategy-startup-period)). This value is used by Freqtrade to ensure that a sufficient amount of data is provided when calling the `dataprovider`, to avoid any NaNs at the beginning of the first training. You can easily set this value by identifying the longest period (in candle units) which is passed to the indicator creation functions (e.g., TA-Lib functions). In the presented example, `startup_candle_count` is 20 since this is the maximum value in `indicators_periods_candles`.
|
||||
|
||||
!!! Note
|
||||
There are instances where the TA-Lib functions actually require more data than just the passed `period` or else the feature dataset gets populated with NaNs. Anecdotally, multiplying the `startup_candle_count` by 2 always leads to a fully NaN free training dataset. Hence, it is typically safest to multiply the expected `startup_candle_count` by 2. Look out for this log message to confirm that the data is clean:
|
||||
|
||||
```
|
||||
2022-08-31 15:14:04 - freqtrade.freqai.data_kitchen - INFO - dropped 0 training points due to NaNs in populated dataset 4319.
|
||||
```
|
||||
|
||||
## Creating a dynamic target threshold
|
||||
|
||||
Deciding when to enter or exit a trade can be done in a dynamic way to reflect current market conditions. FreqAI allows you to return additional information from the training of a model (more info [here](freqai-feature-engineering.md#returning-additional-info-from-training)). For example, the `&*_std/mean` return values describe the statistical distribution of the target/label *during the most recent training*. Comparing a given prediction to these values allows you to know the rarity of the prediction. In `templates/FreqaiExampleStrategy.py`, the `target_roi` and `sell_roi` are defined to be 1.25 z-scores away from the mean which causes predictions that are closer to the mean to be filtered out.
|
||||
|
||||
```python
|
||||
dataframe["target_roi"] = dataframe["&-s_close_mean"] + dataframe["&-s_close_std"] * 1.25
|
||||
dataframe["sell_roi"] = dataframe["&-s_close_mean"] - dataframe["&-s_close_std"] * 1.25
|
||||
```
|
||||
|
||||
To consider the population of *historical predictions* for creating the dynamic target instead of information from the training as discussed above, you would set `fit_live_predictions_candles` in the config to the number of historical prediction candles you wish to use to generate target statistics.
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"fit_live_predictions_candles": 300,
|
||||
}
|
||||
```
|
||||
|
||||
If this value is set, FreqAI will initially use the predictions from the training data and subsequently begin introducing real prediction data as it is generated. FreqAI will save this historical data to be reloaded if you stop and restart a model with the same `identifier`.
|
||||
|
||||
## Using different prediction models
|
||||
|
||||
FreqAI has multiple example prediction model libraries that are ready to be used as is via the flag `--freqaimodel`. These libraries include `CatBoost`, `LightGBM`, and `XGBoost` regression, classification, and multi-target models, and can be found in `freqai/prediction_models/`.
|
||||
|
||||
Regression and classification models differ in what targets they predict - a regression model will predict a target of continuous values, for example what price BTC will be at tomorrow, whilst a classifier will predict a target of discrete values, for example if the price of BTC will go up tomorrow or not. This means that you have to specify your targets differently depending on which model type you are using (see details [below](#setting-model-targets)).
|
||||
|
||||
All of the aforementioned model libraries implement gradient boosted decision tree algorithms. They all work on the principle of ensemble learning, where predictions from multiple simple learners are combined to get a final prediction that is more stable and generalized. The simple learners in this case are decision trees. Gradient boosting refers to the method of learning, where each simple learner is built in sequence - the subsequent learner is used to improve on the error from the previous learner. If you want to learn more about the different model libraries you can find the information in their respective docs:
|
||||
|
||||
* CatBoost: https://catboost.ai/en/docs/
|
||||
* LightGBM: https://lightgbm.readthedocs.io/en/v3.3.2/#
|
||||
* XGBoost: https://xgboost.readthedocs.io/en/stable/#
|
||||
|
||||
There are also numerous online articles describing and comparing the algorithms. Some relatively lightweight examples would be [CatBoost vs. LightGBM vs. XGBoost — Which is the best algorithm?](https://towardsdatascience.com/catboost-vs-lightgbm-vs-xgboost-c80f40662924#:~:text=In%20CatBoost%2C%20symmetric%20trees%2C%20or,the%20same%20depth%20can%20differ.) and [XGBoost, LightGBM or CatBoost — which boosting algorithm should I use?](https://medium.com/riskified-technology/xgboost-lightgbm-or-catboost-which-boosting-algorithm-should-i-use-e7fda7bb36bc). Keep in mind that the performance of each model is highly dependent on the application and so any reported metrics might not be true for your particular use of the model.
|
||||
|
||||
Apart from the models already available in FreqAI, it is also possible to customize and create your own prediction models using the `IFreqaiModel` class. You are encouraged to inherit `fit()`, `train()`, and `predict()` to customize various aspects of the training procedures. You can place custom FreqAI models in `user_data/freqaimodels` - and freqtrade will pick them up from there based on the provided `--freqaimodel` name - which has to correspond to the class name of your custom model.
|
||||
Make sure to use unique names to avoid overriding built-in models.
|
||||
|
||||
### Setting model targets
|
||||
|
||||
#### Regressors
|
||||
|
||||
If you are using a regressor, you need to specify a target that has continuous values. FreqAI includes a variety of regressors, such as the `CatboostRegressor`via the flag `--freqaimodel CatboostRegressor`. An example of how you could set a regression target for predicting the price 100 candles into the future would be
|
||||
|
||||
```python
|
||||
df['&s-close_price'] = df['close'].shift(-100)
|
||||
```
|
||||
|
||||
If you want to predict multiple targets, you need to define multiple labels using the same syntax as shown above.
|
||||
|
||||
#### Classifiers
|
||||
|
||||
If you are using a classifier, you need to specify a target that has discrete values. FreqAI includes a variety of classifiers, such as the `CatboostClassifier` via the flag `--freqaimodel CatboostClassifier`. If you elects to use a classifier, the classes need to be set using strings. For example, if you want to predict if the price 100 candles into the future goes up or down you would set
|
||||
|
||||
```python
|
||||
df['&s-up_or_down'] = np.where( df["close"].shift(-100) > df["close"], 'up', 'down')
|
||||
```
|
||||
|
||||
If you want to predict multiple targets you must specify all labels in the same label column. You could, for example, add the label `same` to define where the price was unchanged by setting
|
||||
|
||||
```python
|
||||
df['&s-up_or_down'] = np.where( df["close"].shift(-100) > df["close"], 'up', 'down')
|
||||
df['&s-up_or_down'] = np.where( df["close"].shift(-100) == df["close"], 'same', df['&s-up_or_down'])
|
||||
```
|
||||
|
||||
## PyTorch Module
|
||||
|
||||
### Quick start
|
||||
|
||||
The easiest way to quickly run a pytorch model is with the following command (for regression task):
|
||||
|
||||
```bash
|
||||
freqtrade trade --config config_examples/config_freqai.example.json --strategy FreqaiExampleStrategy --freqaimodel PyTorchMLPRegressor --strategy-path freqtrade/templates
|
||||
```
|
||||
|
||||
!!! Note "Installation/docker"
|
||||
The PyTorch module requires large packages such as `torch`, which should be explicitly requested during `./setup.sh -i` by answering "y" to the question "Do you also want dependencies for freqai-rl or PyTorch (~700mb additional space required) [y/N]?".
|
||||
Users who prefer docker should ensure they use the docker image appended with `_freqaitorch`.
|
||||
We do provide an explicit docker-compose file for this in `docker/docker-compose-freqai.yml` - which can be used via `docker compose -f docker/docker-compose-freqai.yml run ...` - or can be copied to replace the original docker file.
|
||||
This docker-compose file also contains a (disabled) section to enable GPU resources within docker containers. This obviously assumes the system has GPU resources available.
|
||||
|
||||
PyTorch dropped support for macOS x64 (intel based Apple devices) in version 2.3. Subsequently, freqtrade also dropped support for PyTorch on this platform.
|
||||
|
||||
### Structure
|
||||
|
||||
#### Model
|
||||
|
||||
You can construct your own Neural Network architecture in PyTorch by simply defining your `nn.Module` class inside your custom [`IFreqaiModel` file](#using-different-prediction-models) and then using that class in your `def train()` function. Here is an example of logistic regression model implementation using PyTorch (should be used with nn.BCELoss criterion) for classification tasks.
|
||||
|
||||
```python
|
||||
|
||||
class LogisticRegression(nn.Module):
|
||||
def __init__(self, input_size: int):
|
||||
super().__init__()
|
||||
# Define your layers
|
||||
self.linear = nn.Linear(input_size, 1)
|
||||
self.activation = nn.Sigmoid()
|
||||
|
||||
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||||
# Define the forward pass
|
||||
out = self.linear(x)
|
||||
out = self.activation(out)
|
||||
return out
|
||||
|
||||
class MyCoolPyTorchClassifier(BasePyTorchClassifier):
|
||||
"""
|
||||
This is a custom IFreqaiModel showing how a user might setup their own
|
||||
custom Neural Network architecture for their training.
|
||||
"""
|
||||
|
||||
@property
|
||||
def data_convertor(self) -> PyTorchDataConvertor:
|
||||
return DefaultPyTorchDataConvertor(target_tensor_type=torch.float)
|
||||
|
||||
def __init__(self, **kwargs) -> None:
|
||||
super().__init__(**kwargs)
|
||||
config = self.freqai_info.get("model_training_parameters", {})
|
||||
self.learning_rate: float = config.get("learning_rate", 3e-4)
|
||||
self.model_kwargs: dict[str, Any] = config.get("model_kwargs", {})
|
||||
self.trainer_kwargs: dict[str, Any] = config.get("trainer_kwargs", {})
|
||||
|
||||
def fit(self, data_dictionary: dict, dk: FreqaiDataKitchen, **kwargs) -> Any:
|
||||
"""
|
||||
User sets up the training and test data to fit their desired model here
|
||||
:param data_dictionary: the dictionary holding all data for train, test,
|
||||
labels, weights
|
||||
:param dk: The datakitchen object for the current coin/model
|
||||
"""
|
||||
|
||||
class_names = self.get_class_names()
|
||||
self.convert_label_column_to_int(data_dictionary, dk, class_names)
|
||||
n_features = data_dictionary["train_features"].shape[-1]
|
||||
model = LogisticRegression(
|
||||
input_dim=n_features
|
||||
)
|
||||
model.to(self.device)
|
||||
optimizer = torch.optim.AdamW(model.parameters(), lr=self.learning_rate)
|
||||
criterion = torch.nn.CrossEntropyLoss()
|
||||
init_model = self.get_init_model(dk.pair)
|
||||
trainer = PyTorchModelTrainer(
|
||||
model=model,
|
||||
optimizer=optimizer,
|
||||
criterion=criterion,
|
||||
model_meta_data={"class_names": class_names},
|
||||
device=self.device,
|
||||
init_model=init_model,
|
||||
data_convertor=self.data_convertor,
|
||||
**self.trainer_kwargs,
|
||||
)
|
||||
trainer.fit(data_dictionary, self.splits)
|
||||
return trainer
|
||||
|
||||
```
|
||||
|
||||
#### Trainer
|
||||
|
||||
The `PyTorchModelTrainer` performs the idiomatic PyTorch train loop:
|
||||
Define our model, loss function, and optimizer, and then move them to the appropriate device (GPU or CPU). Inside the loop, we iterate through the batches in the dataloader, move the data to the device, compute the prediction and loss, backpropagate, and update the model parameters using the optimizer.
|
||||
|
||||
In addition, the trainer is responsible for the following:
|
||||
- saving and loading the model
|
||||
- converting the data from `pandas.DataFrame` to `torch.Tensor`.
|
||||
|
||||
#### Integration with Freqai module
|
||||
|
||||
Like all freqai models, PyTorch models inherit `IFreqaiModel`. `IFreqaiModel` declares three abstract methods: `train`, `fit`, and `predict`. we implement these methods in three levels of hierarchy.
|
||||
From top to bottom:
|
||||
|
||||
1. `BasePyTorchModel` - Implements the `train` method. all `BasePyTorch*` inherit it. responsible for general data preparation (e.g., data normalization) and calling the `fit` method. Sets `device` attribute used by children classes. Sets `model_type` attribute used by the parent class.
|
||||
2. `BasePyTorch*` - Implements the `predict` method. Here, the `*` represents a group of algorithms, such as classifiers or regressors. responsible for data preprocessing, predicting, and postprocessing if needed.
|
||||
3. `PyTorch*Classifier` / `PyTorch*Regressor` - implements the `fit` method. responsible for the main train flaw, where we initialize the trainer and model objects.
|
||||
|
||||
|
||||
|
||||
#### Full example
|
||||
|
||||
Building a PyTorch regressor using MLP (multilayer perceptron) model, MSELoss criterion, and AdamW optimizer.
|
||||
|
||||
```python
|
||||
class PyTorchMLPRegressor(BasePyTorchRegressor):
|
||||
def __init__(self, **kwargs) -> None:
|
||||
super().__init__(**kwargs)
|
||||
config = self.freqai_info.get("model_training_parameters", {})
|
||||
self.learning_rate: float = config.get("learning_rate", 3e-4)
|
||||
self.model_kwargs: dict[str, Any] = config.get("model_kwargs", {})
|
||||
self.trainer_kwargs: dict[str, Any] = config.get("trainer_kwargs", {})
|
||||
|
||||
def fit(self, data_dictionary: dict, dk: FreqaiDataKitchen, **kwargs) -> Any:
|
||||
n_features = data_dictionary["train_features"].shape[-1]
|
||||
model = PyTorchMLPModel(
|
||||
input_dim=n_features,
|
||||
output_dim=1,
|
||||
**self.model_kwargs
|
||||
)
|
||||
model.to(self.device)
|
||||
optimizer = torch.optim.AdamW(model.parameters(), lr=self.learning_rate)
|
||||
criterion = torch.nn.MSELoss()
|
||||
init_model = self.get_init_model(dk.pair)
|
||||
trainer = PyTorchModelTrainer(
|
||||
model=model,
|
||||
optimizer=optimizer,
|
||||
criterion=criterion,
|
||||
device=self.device,
|
||||
init_model=init_model,
|
||||
target_tensor_type=torch.float,
|
||||
**self.trainer_kwargs,
|
||||
)
|
||||
trainer.fit(data_dictionary)
|
||||
return trainer
|
||||
```
|
||||
|
||||
Here we create a `PyTorchMLPRegressor` class that implements the `fit` method. The `fit` method specifies the training building blocks: model, optimizer, criterion, and trainer. We inherit both `BasePyTorchRegressor` and `BasePyTorchModel`, where the former implements the `predict` method that is suitable for our regression task, and the latter implements the train method.
|
||||
|
||||
??? Note "Setting Class Names for Classifiers"
|
||||
When using classifiers, the user must declare the class names (or targets) by overriding the `IFreqaiModel.class_names` attribute. This is achieved by setting `self.freqai.class_names` in the FreqAI strategy inside the `set_freqai_targets` method.
|
||||
|
||||
For example, if you are using a binary classifier to predict price movements as up or down, you can set the class names as follows:
|
||||
```python
|
||||
def set_freqai_targets(self, dataframe: DataFrame, metadata: dict, **kwargs) -> DataFrame:
|
||||
self.freqai.class_names = ["down", "up"]
|
||||
dataframe['&s-up_or_down'] = np.where(dataframe["close"].shift(-100) >
|
||||
dataframe["close"], 'up', 'down')
|
||||
|
||||
return dataframe
|
||||
```
|
||||
To see a full example, you can refer to the [classifier test strategy class](https://github.com/freqtrade/freqtrade/blob/develop/tests/strategy/strats/freqai_test_classifier.py).
|
||||
|
||||
|
||||
#### Improving performance with `torch.compile()`
|
||||
|
||||
Torch provides a `torch.compile()` method that can be used to improve performance for specific GPU hardware. More details can be found [here](https://pytorch.org/tutorials/intermediate/torch_compile_tutorial.html). In brief, you simply wrap your `model` in `torch.compile()`:
|
||||
|
||||
|
||||
```python
|
||||
model = PyTorchMLPModel(
|
||||
input_dim=n_features,
|
||||
output_dim=1,
|
||||
**self.model_kwargs
|
||||
)
|
||||
model.to(self.device)
|
||||
model = torch.compile(model)
|
||||
```
|
||||
|
||||
Then proceed to use the model as normal. Keep in mind that doing this will remove eager execution, which means errors and tracebacks will not be informative.
|
||||
78
md/freqai-developers.md
Normal file
78
md/freqai-developers.md
Normal file
@ -0,0 +1,78 @@
|
||||
# Development
|
||||
|
||||
## Project architecture
|
||||
|
||||
The architecture and functions of FreqAI are generalized to encourages development of unique features, functions, models, etc.
|
||||
|
||||
The class structure and a detailed algorithmic overview is depicted in the following diagram:
|
||||
|
||||
|
||||
|
||||
As shown, there are three distinct objects comprising FreqAI:
|
||||
|
||||
* **IFreqaiModel** - A singular persistent object containing all the necessary logic to collect, store, and process data, engineer features, run training, and inference models.
|
||||
* **FreqaiDataKitchen** - A non-persistent object which is created uniquely for each unique asset/model. Beyond metadata, it also contains a variety of data processing tools.
|
||||
* **FreqaiDataDrawer** - A singular persistent object containing all the historical predictions, models, and save/load methods.
|
||||
|
||||
There are a variety of built-in [prediction models](freqai-configuration.md#using-different-prediction-models) which inherit directly from `IFreqaiModel`. Each of these models have full access to all methods in `IFreqaiModel` and can therefore override any of those functions at will. However, advanced users will likely stick to overriding `fit()`, `train()`, `predict()`, and `data_cleaning_train/predict()`.
|
||||
|
||||
## Data handling
|
||||
|
||||
FreqAI aims to organize model files, prediction data, and meta data in a way that simplifies post-processing and enhances crash resilience by automatic data reloading. The data is saved in a file structure,`user_data_dir/models/`, which contains all the data associated with the trainings and backtests. The `FreqaiDataKitchen()` relies heavily on the file structure for proper training and inferencing and should therefore not be manually modified.
|
||||
|
||||
### File structure
|
||||
|
||||
The file structure is automatically generated based on the model `identifier` set in the [config](freqai-configuration.md#setting-up-the-configuration-file). The following structure shows where the data is stored for post processing:
|
||||
|
||||
| Structure | Description |
|
||||
|-----------|-------------|
|
||||
| `config_*.json` | A copy of the model specific configuration file. |
|
||||
| `historic_predictions.pkl` | A file containing all historic predictions generated during the lifetime of the `identifier` model during live deployment. `historic_predictions.pkl` is used to reload the model after a crash or a config change. A backup file is always held in case of corruption on the main file. FreqAI **automatically** detects corruption and replaces the corrupted file with the backup. |
|
||||
| `pair_dictionary.json` | A file containing the training queue as well as the on disk location of the most recently trained model. |
|
||||
| `sub-train-*_TIMESTAMP` | A folder containing all the files associated with a single model, such as: <br>
|
||||
|| `*_metadata.json` - Metadata for the model, such as normalization max/min, expected training feature list, etc. <br>
|
||||
|| `*_model.*` - The model file saved to disk for reloading from a crash. Can be `joblib` (typical boosting libs), `zip` (stable_baselines), `hd5` (keras type), etc. <br>
|
||||
|| `*_pca_object.pkl` - The [Principal component analysis (PCA)](freqai-feature-engineering.md#data-dimensionality-reduction-with-principal-component-analysis) transform (if `principal_component_analysis: True` is set in the config) which will be used to transform unseen prediction features. <br>
|
||||
|| `*_svm_model.pkl` - The [Support Vector Machine (SVM)](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm) model (if `use_SVM_to_remove_outliers: True` is set in the config) which is used to detect outliers in unseen prediction features. <br>
|
||||
|| `*_trained_df.pkl` - The dataframe containing all the training features used to train the `identifier` model. This is used for computing the [Dissimilarity Index (DI)](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di) and can also be used for post-processing. <br>
|
||||
|| `*_trained_dates.df.pkl` - The dates associated with the `trained_df.pkl`, which is useful for post-processing. |
|
||||
|
||||
The example file structure would look like this:
|
||||
|
||||
```
|
||||
├── models
|
||||
│ └── unique-id
|
||||
│ ├── config_freqai.example.json
|
||||
│ ├── historic_predictions.backup.pkl
|
||||
│ ├── historic_predictions.pkl
|
||||
│ ├── pair_dictionary.json
|
||||
│ ├── sub-train-1INCH_1662821319
|
||||
│ │ ├── cb_1inch_1662821319_metadata.json
|
||||
│ │ ├── cb_1inch_1662821319_model.joblib
|
||||
│ │ ├── cb_1inch_1662821319_pca_object.pkl
|
||||
│ │ ├── cb_1inch_1662821319_svm_model.joblib
|
||||
│ │ ├── cb_1inch_1662821319_trained_dates_df.pkl
|
||||
│ │ └── cb_1inch_1662821319_trained_df.pkl
|
||||
│ ├── sub-train-1INCH_1662821371
|
||||
│ │ ├── cb_1inch_1662821371_metadata.json
|
||||
│ │ ├── cb_1inch_1662821371_model.joblib
|
||||
│ │ ├── cb_1inch_1662821371_pca_object.pkl
|
||||
│ │ ├── cb_1inch_1662821371_svm_model.joblib
|
||||
│ │ ├── cb_1inch_1662821371_trained_dates_df.pkl
|
||||
│ │ └── cb_1inch_1662821371_trained_df.pkl
|
||||
│ ├── sub-train-ADA_1662821344
|
||||
│ │ ├── cb_ada_1662821344_metadata.json
|
||||
│ │ ├── cb_ada_1662821344_model.joblib
|
||||
│ │ ├── cb_ada_1662821344_pca_object.pkl
|
||||
│ │ ├── cb_ada_1662821344_svm_model.joblib
|
||||
│ │ ├── cb_ada_1662821344_trained_dates_df.pkl
|
||||
│ │ └── cb_ada_1662821344_trained_df.pkl
|
||||
│ └── sub-train-ADA_1662821399
|
||||
│ ├── cb_ada_1662821399_metadata.json
|
||||
│ ├── cb_ada_1662821399_model.joblib
|
||||
│ ├── cb_ada_1662821399_pca_object.pkl
|
||||
│ ├── cb_ada_1662821399_svm_model.joblib
|
||||
│ ├── cb_ada_1662821399_trained_dates_df.pkl
|
||||
│ └── cb_ada_1662821399_trained_df.pkl
|
||||
|
||||
```
|
||||
408
md/freqai-feature-engineering.md
Normal file
408
md/freqai-feature-engineering.md
Normal file
@ -0,0 +1,408 @@
|
||||
# Feature engineering
|
||||
|
||||
## Defining the features
|
||||
|
||||
Low level feature engineering is performed in the user strategy within a set of functions called `feature_engineering_*`. These function set the `base features` such as, `RSI`, `MFI`, `EMA`, `SMA`, time of day, volume, etc. The `base features` can be custom indicators or they can be imported from any technical-analysis library that you can find. FreqAI is equipped with a set of functions to simplify rapid large-scale feature engineering:
|
||||
|
||||
| Function | Description |
|
||||
|---------------|-------------|
|
||||
| `feature_engineering_expand_all()` | This optional function will automatically expand the defined features on the config defined `indicator_periods_candles`, `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`.
|
||||
| `feature_engineering_expand_basic()` | This optional function will automatically expand the defined features on the config defined `include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`. Note: this function does *not* expand across `indicator_periods_candles`.
|
||||
| `feature_engineering_standard()` | This optional function will be called once with the dataframe of the base timeframe. This is the final function to be called, which means that the dataframe entering this function will contain all the features and columns from the base asset created by the other `feature_engineering_expand` functions. This function is a good place to do custom exotic feature extractions (e.g. tsfresh). This function is also a good place for any feature that should not be auto-expanded upon (e.g., day of the week).
|
||||
| `set_freqai_targets()` | Required function to set the targets for the model. All targets must be prepended with `&` to be recognized by the FreqAI internals.
|
||||
|
||||
Meanwhile, high level feature engineering is handled within `"feature_parameters":{}` in the FreqAI config. Within this file, it is possible to decide large scale feature expansions on top of the `base_features` such as "including correlated pairs" or "including informative timeframes" or even "including recent candles."
|
||||
|
||||
It is advisable to start from the template `feature_engineering_*` functions in the source provided example strategy (found in `templates/FreqaiExampleStrategy.py`) to ensure that the feature definitions are following the correct conventions. Here is an example of how to set the indicators and labels in the strategy:
|
||||
|
||||
```python
|
||||
def feature_engineering_expand_all(self, dataframe: DataFrame, period, metadata, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This function will automatically expand the defined features on the config defined
|
||||
`indicator_periods_candles`, `include_timeframes`, `include_shifted_candles`, and
|
||||
`include_corr_pairs`. In other words, a single feature defined in this function
|
||||
will automatically expand to a total of
|
||||
`indicator_periods_candles` * `include_timeframes` * `include_shifted_candles` *
|
||||
`include_corr_pairs` numbers of features added to the model.
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
Access metadata such as the current pair/timeframe/period with:
|
||||
|
||||
`metadata["pair"]` `metadata["tf"]` `metadata["period"]`
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
:param period: period of the indicator - usage example:
|
||||
:param metadata: metadata of current pair
|
||||
dataframe["%-ema-period"] = ta.EMA(dataframe, timeperiod=period)
|
||||
"""
|
||||
|
||||
dataframe["%-rsi-period"] = ta.RSI(dataframe, timeperiod=period)
|
||||
dataframe["%-mfi-period"] = ta.MFI(dataframe, timeperiod=period)
|
||||
dataframe["%-adx-period"] = ta.ADX(dataframe, timeperiod=period)
|
||||
dataframe["%-sma-period"] = ta.SMA(dataframe, timeperiod=period)
|
||||
dataframe["%-ema-period"] = ta.EMA(dataframe, timeperiod=period)
|
||||
|
||||
bollinger = qtpylib.bollinger_bands(
|
||||
qtpylib.typical_price(dataframe), window=period, stds=2.2
|
||||
)
|
||||
dataframe["bb_lowerband-period"] = bollinger["lower"]
|
||||
dataframe["bb_middleband-period"] = bollinger["mid"]
|
||||
dataframe["bb_upperband-period"] = bollinger["upper"]
|
||||
|
||||
dataframe["%-bb_width-period"] = (
|
||||
dataframe["bb_upperband-period"]
|
||||
- dataframe["bb_lowerband-period"]
|
||||
) / dataframe["bb_middleband-period"]
|
||||
dataframe["%-close-bb_lower-period"] = (
|
||||
dataframe["close"] / dataframe["bb_lowerband-period"]
|
||||
)
|
||||
|
||||
dataframe["%-roc-period"] = ta.ROC(dataframe, timeperiod=period)
|
||||
|
||||
dataframe["%-relative_volume-period"] = (
|
||||
dataframe["volume"] / dataframe["volume"].rolling(period).mean()
|
||||
)
|
||||
|
||||
return dataframe
|
||||
|
||||
def feature_engineering_expand_basic(self, dataframe: DataFrame, metadata, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This function will automatically expand the defined features on the config defined
|
||||
`include_timeframes`, `include_shifted_candles`, and `include_corr_pairs`.
|
||||
In other words, a single feature defined in this function
|
||||
will automatically expand to a total of
|
||||
`include_timeframes` * `include_shifted_candles` * `include_corr_pairs`
|
||||
numbers of features added to the model.
|
||||
|
||||
Features defined here will *not* be automatically duplicated on user defined
|
||||
`indicator_periods_candles`
|
||||
|
||||
Access metadata such as the current pair/timeframe with:
|
||||
|
||||
`metadata["pair"]` `metadata["tf"]`
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
:param metadata: metadata of current pair
|
||||
dataframe["%-pct-change"] = dataframe["close"].pct_change()
|
||||
dataframe["%-ema-200"] = ta.EMA(dataframe, timeperiod=200)
|
||||
"""
|
||||
dataframe["%-pct-change"] = dataframe["close"].pct_change()
|
||||
dataframe["%-raw_volume"] = dataframe["volume"]
|
||||
dataframe["%-raw_price"] = dataframe["close"]
|
||||
return dataframe
|
||||
|
||||
def feature_engineering_standard(self, dataframe: DataFrame, metadata, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
This optional function will be called once with the dataframe of the base timeframe.
|
||||
This is the final function to be called, which means that the dataframe entering this
|
||||
function will contain all the features and columns created by all other
|
||||
freqai_feature_engineering_* functions.
|
||||
|
||||
This function is a good place to do custom exotic feature extractions (e.g. tsfresh).
|
||||
This function is a good place for any feature that should not be auto-expanded upon
|
||||
(e.g. day of the week).
|
||||
|
||||
Access metadata such as the current pair with:
|
||||
|
||||
`metadata["pair"]`
|
||||
|
||||
All features must be prepended with `%` to be recognized by FreqAI internals.
|
||||
|
||||
:param df: strategy dataframe which will receive the features
|
||||
:param metadata: metadata of current pair
|
||||
usage example: dataframe["%-day_of_week"] = (dataframe["date"].dt.dayofweek + 1) / 7
|
||||
"""
|
||||
dataframe["%-day_of_week"] = (dataframe["date"].dt.dayofweek + 1) / 7
|
||||
dataframe["%-hour_of_day"] = (dataframe["date"].dt.hour + 1) / 25
|
||||
return dataframe
|
||||
|
||||
def set_freqai_targets(self, dataframe: DataFrame, metadata, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
Required function to set the targets for the model.
|
||||
All targets must be prepended with `&` to be recognized by the FreqAI internals.
|
||||
|
||||
Access metadata such as the current pair with:
|
||||
|
||||
`metadata["pair"]`
|
||||
|
||||
:param df: strategy dataframe which will receive the targets
|
||||
:param metadata: metadata of current pair
|
||||
usage example: dataframe["&-target"] = dataframe["close"].shift(-1) / dataframe["close"]
|
||||
"""
|
||||
dataframe["&-s_close"] = (
|
||||
dataframe["close"]
|
||||
.shift(-self.freqai_info["feature_parameters"]["label_period_candles"])
|
||||
.rolling(self.freqai_info["feature_parameters"]["label_period_candles"])
|
||||
.mean()
|
||||
/ dataframe["close"]
|
||||
- 1
|
||||
)
|
||||
|
||||
return dataframe
|
||||
```
|
||||
|
||||
In the presented example, the user does not wish to pass the `bb_lowerband` as a feature to the model,
|
||||
and has therefore not prepended it with `%`. The user does, however, wish to pass `bb_width` to the
|
||||
model for training/prediction and has therefore prepended it with `%`.
|
||||
|
||||
After having defined the `base features`, the next step is to expand upon them using the powerful `feature_parameters` in the configuration file:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
//...
|
||||
"feature_parameters" : {
|
||||
"include_timeframes": ["5m","15m","4h"],
|
||||
"include_corr_pairlist": [
|
||||
"ETH/USD",
|
||||
"LINK/USD",
|
||||
"BNB/USD"
|
||||
],
|
||||
"label_period_candles": 24,
|
||||
"include_shifted_candles": 2,
|
||||
"indicator_periods_candles": [10, 20]
|
||||
},
|
||||
//...
|
||||
}
|
||||
```
|
||||
|
||||
The `include_timeframes` in the config above are the timeframes (`tf`) of each call to `feature_engineering_expand_*()` in the strategy. In the presented case, the user is asking for the `5m`, `15m`, and `4h` timeframes of the `rsi`, `mfi`, `roc`, and `bb_width` to be included in the feature set.
|
||||
|
||||
You can ask for each of the defined features to be included also for informative pairs using the `include_corr_pairlist`. This means that the feature set will include all the features from `feature_engineering_expand_*()` on all the `include_timeframes` for each of the correlated pairs defined in the config (`ETH/USD`, `LINK/USD`, and `BNB/USD` in the presented example).
|
||||
|
||||
`include_shifted_candles` indicates the number of previous candles to include in the feature set. For example, `include_shifted_candles: 2` tells FreqAI to include the past 2 candles for each of the features in the feature set.
|
||||
|
||||
In total, the number of features the user of the presented example strategy has created is: length of `include_timeframes` * no. features in `feature_engineering_expand_*()` * length of `include_corr_pairlist` * no. `include_shifted_candles` * length of `indicator_periods_candles`
|
||||
$= 3 * 3 * 3 * 2 * 2 = 108$.
|
||||
|
||||
!!! note "Learn more about creative feature engineering"
|
||||
Check out our [medium article](https://emergentmethods.medium.com/freqai-from-price-to-prediction-6fadac18b665) geared toward helping users learn how to creatively engineer features.
|
||||
|
||||
### Gain finer control over `feature_engineering_*` functions with `metadata`
|
||||
|
||||
All `feature_engineering_*` and `set_freqai_targets()` functions are passed a `metadata` dictionary which contains information about the `pair`, `tf` (timeframe), and `period` that FreqAI is automating for feature building. As such, a user can use `metadata` inside `feature_engineering_*` functions as criteria for blocking/reserving features for certain timeframes, periods, pairs etc.
|
||||
|
||||
```python
|
||||
def feature_engineering_expand_all(self, dataframe: DataFrame, period, metadata, **kwargs) -> DataFrame:
|
||||
if metadata["tf"] == "1h":
|
||||
dataframe["%-roc-period"] = ta.ROC(dataframe, timeperiod=period)
|
||||
```
|
||||
|
||||
This will block `ta.ROC()` from being added to any timeframes other than `"1h"`.
|
||||
|
||||
### Returning additional info from training
|
||||
|
||||
Important metrics can be returned to the strategy at the end of each model training by assigning them to `dk.data['extra_returns_per_train']['my_new_value'] = XYZ` inside the custom prediction model class.
|
||||
|
||||
FreqAI takes the `my_new_value` assigned in this dictionary and expands it to fit the dataframe that is returned to the strategy. You can then use the returned metrics in your strategy through `dataframe['my_new_value']`. An example of how return values can be used in FreqAI are the `&*_mean` and `&*_std` values that are used to [created a dynamic target threshold](freqai-configuration.md#creating-a-dynamic-target-threshold).
|
||||
|
||||
Another example, where the user wants to use live metrics from the trade database, is shown below:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"extra_returns_per_train": {"total_profit": 4}
|
||||
}
|
||||
```
|
||||
|
||||
You need to set the standard dictionary in the config so that FreqAI can return proper dataframe shapes. These values will likely be overridden by the prediction model, but in the case where the model has yet to set them, or needs a default initial value, the pre-set values are what will be returned.
|
||||
|
||||
### Weighting features for temporal importance
|
||||
|
||||
FreqAI allows you to set a `weight_factor` to weight recent data more strongly than past data via an exponential function:
|
||||
|
||||
$$ W_i = \exp(\frac{-i}{\alpha*n}) $$
|
||||
|
||||
where $W_i$ is the weight of data point $i$ in a total set of $n$ data points. Below is a figure showing the effect of different weight factors on the data points in a feature set.
|
||||
|
||||
|
||||
|
||||
## Building the data pipeline
|
||||
|
||||
By default, FreqAI builds a dynamic pipeline based on user configuration settings. The default settings are robust and designed to work with a variety of methods. These two steps are a `MinMaxScaler(-1,1)` and a `VarianceThreshold` which removes any column that has 0 variance. Users can activate other steps with more configuration parameters. For example if users add `use_SVM_to_remove_outliers: true` to the `freqai` config, then FreqAI will automatically add the [`SVMOutlierExtractor`](#identifying-outliers-using-a-support-vector-machine-svm) to the pipeline. Likewise, users can add `principal_component_analysis: true` to the `freqai` config to activate PCA. The [DissimilarityIndex](#identifying-outliers-with-the-dissimilarity-index-di) is activated with `DI_threshold: 1`. Finally, noise can also be added to the data with `noise_standard_deviation: 0.1`. Finally, users can add [DBSCAN](#identifying-outliers-with-dbscan) outlier removal with `use_DBSCAN_to_remove_outliers: true`.
|
||||
|
||||
!!! note "More information available"
|
||||
Please review the [parameter table](freqai-parameter-table.md) for more information on these parameters.
|
||||
|
||||
|
||||
### Customizing the pipeline
|
||||
|
||||
Users are encouraged to customize the data pipeline to their needs by building their own data pipeline. This can be done by simply setting `dk.feature_pipeline` to their desired `Pipeline` object inside their `IFreqaiModel` `train()` function, or if they prefer not to touch the `train()` function, they can override `define_data_pipeline`/`define_label_pipeline` functions in their `IFreqaiModel`:
|
||||
|
||||
!!! note "More information available"
|
||||
FreqAI uses the [`DataSieve`](https://github.com/emergentmethods/datasieve) pipeline, which follows the SKlearn pipeline API, but adds, among other features, coherence between the X, y, and sample_weight vector point removals, feature removal, feature name following.
|
||||
|
||||
```python
|
||||
from datasieve.transforms import SKLearnWrapper, DissimilarityIndex
|
||||
from datasieve.pipeline import Pipeline
|
||||
from sklearn.preprocessing import QuantileTransformer, StandardScaler
|
||||
from freqai.base_models import BaseRegressionModel
|
||||
|
||||
|
||||
class MyFreqaiModel(BaseRegressionModel):
|
||||
"""
|
||||
Some cool custom model
|
||||
"""
|
||||
def fit(self, data_dictionary: Dict, dk: FreqaiDataKitchen, **kwargs) -> Any:
|
||||
"""
|
||||
My custom fit function
|
||||
"""
|
||||
model = cool_model.fit()
|
||||
return model
|
||||
|
||||
def define_data_pipeline(self) -> Pipeline:
|
||||
"""
|
||||
User defines their custom feature pipeline here (if they wish)
|
||||
"""
|
||||
feature_pipeline = Pipeline([
|
||||
('qt', SKLearnWrapper(QuantileTransformer(output_distribution='normal'))),
|
||||
('di', ds.DissimilarityIndex(di_threshold=1))
|
||||
])
|
||||
|
||||
return feature_pipeline
|
||||
|
||||
def define_label_pipeline(self) -> Pipeline:
|
||||
"""
|
||||
User defines their custom label pipeline here (if they wish)
|
||||
"""
|
||||
label_pipeline = Pipeline([
|
||||
('qt', SKLearnWrapper(StandardScaler())),
|
||||
])
|
||||
|
||||
return label_pipeline
|
||||
```
|
||||
|
||||
Here, you are defining the exact pipeline that will be used for your feature set during training and prediction. You can use *most* SKLearn transformation steps by wrapping them in the `SKLearnWrapper` class as shown above. In addition, you can use any of the transformations available in the [`DataSieve` library](https://github.com/emergentmethods/datasieve).
|
||||
|
||||
You can easily add your own transformation by creating a class that inherits from the datasieve `BaseTransform` and implementing your `fit()`, `transform()` and `inverse_transform()` methods:
|
||||
|
||||
```python
|
||||
from datasieve.transforms.base_transform import BaseTransform
|
||||
# import whatever else you need
|
||||
|
||||
class MyCoolTransform(BaseTransform):
|
||||
def __init__(self, **kwargs):
|
||||
self.param1 = kwargs.get('param1', 1)
|
||||
|
||||
def fit(self, X, y=None, sample_weight=None, feature_list=None, **kwargs):
|
||||
# do something with X, y, sample_weight, or/and feature_list
|
||||
return X, y, sample_weight, feature_list
|
||||
|
||||
def transform(self, X, y=None, sample_weight=None,
|
||||
feature_list=None, outlier_check=False, **kwargs):
|
||||
# do something with X, y, sample_weight, or/and feature_list
|
||||
return X, y, sample_weight, feature_list
|
||||
|
||||
def inverse_transform(self, X, y=None, sample_weight=None, feature_list=None, **kwargs):
|
||||
# do/dont do something with X, y, sample_weight, or/and feature_list
|
||||
return X, y, sample_weight, feature_list
|
||||
```
|
||||
|
||||
!!! note "Hint"
|
||||
You can define this custom class in the same file as your `IFreqaiModel`.
|
||||
|
||||
### Migrating a custom `IFreqaiModel` to the new Pipeline
|
||||
|
||||
If you have created your own custom `IFreqaiModel` with a custom `train()`/`predict()` function, *and* you still rely on `data_cleaning_train/predict()`, then you will need to migrate to the new pipeline. If your model does *not* rely on `data_cleaning_train/predict()`, then you do not need to worry about this migration.
|
||||
|
||||
More details about the migration can be found [here](strategy_migration.md#freqai---new-data-pipeline).
|
||||
|
||||
## Outlier detection
|
||||
|
||||
Equity and crypto markets suffer from a high level of non-patterned noise in the form of outlier data points. FreqAI implements a variety of methods to identify such outliers and hence mitigate risk.
|
||||
|
||||
### Identifying outliers with the Dissimilarity Index (DI)
|
||||
|
||||
The Dissimilarity Index (DI) aims to quantify the uncertainty associated with each prediction made by the model.
|
||||
|
||||
You can tell FreqAI to remove outlier data points from the training/test data sets using the DI by including the following statement in the config:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"feature_parameters" : {
|
||||
"DI_threshold": 1
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Which will add `DissimilarityIndex` step to your `feature_pipeline` and set the threshold to 1. The DI allows predictions which are outliers (not existent in the model feature space) to be thrown out due to low levels of certainty. To do so, FreqAI measures the distance between each training data point (feature vector), $X_{a}$, and all other training data points:
|
||||
|
||||
$$ d_{ab} = \sqrt{\sum_{j=1}^p(X_{a,j}-X_{b,j})^2} $$
|
||||
|
||||
where $d_{ab}$ is the distance between the normalized points $a$ and $b$, and $p$ is the number of features, i.e., the length of the vector $X$. The characteristic distance, $\overline{d}$, for a set of training data points is simply the mean of the average distances:
|
||||
|
||||
$$ \overline{d} = \sum_{a=1}^n(\sum_{b=1}^n(d_{ab}/n)/n) $$
|
||||
|
||||
$\overline{d}$ quantifies the spread of the training data, which is compared to the distance between a new prediction feature vectors, $X_k$ and all the training data:
|
||||
|
||||
$$ d_k = \arg \min d_{k,i} $$
|
||||
|
||||
This enables the estimation of the Dissimilarity Index as:
|
||||
|
||||
$$ DI_k = d_k/\overline{d} $$
|
||||
|
||||
You can tweak the DI through the `DI_threshold` to increase or decrease the extrapolation of the trained model. A higher `DI_threshold` means that the DI is more lenient and allows predictions further away from the training data to be used whilst a lower `DI_threshold` has the opposite effect and hence discards more predictions.
|
||||
|
||||
Below is a figure that describes the DI for a 3D data set.
|
||||
|
||||
|
||||
|
||||
### Identifying outliers using a Support Vector Machine (SVM)
|
||||
|
||||
You can tell FreqAI to remove outlier data points from the training/test data sets using a Support Vector Machine (SVM) by including the following statement in the config:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"feature_parameters" : {
|
||||
"use_SVM_to_remove_outliers": true
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Which will add `SVMOutlierExtractor` step to your `feature_pipeline`. The SVM will be trained on the training data and any data point that the SVM deems to be beyond the feature space will be removed.
|
||||
|
||||
You can elect to provide additional parameters for the SVM, such as `shuffle`, and `nu` via the `feature_parameters.svm_params` dictionary in the config.
|
||||
|
||||
The parameter `shuffle` is by default set to `False` to ensure consistent results. If it is set to `True`, running the SVM multiple times on the same data set might result in different outcomes due to `max_iter` being to low for the algorithm to reach the demanded `tol`. Increasing `max_iter` solves this issue but causes the procedure to take longer time.
|
||||
|
||||
The parameter `nu`, *very* broadly, is the amount of data points that should be considered outliers and should be between 0 and 1.
|
||||
|
||||
### Identifying outliers with DBSCAN
|
||||
|
||||
You can configure FreqAI to use DBSCAN to cluster and remove outliers from the training/test data set or incoming outliers from predictions, by activating `use_DBSCAN_to_remove_outliers` in the config:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"feature_parameters" : {
|
||||
"use_DBSCAN_to_remove_outliers": true
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Which will add the `DataSieveDBSCAN` step to your `feature_pipeline`. This is an unsupervised machine learning algorithm that clusters data without needing to know how many clusters there should be.
|
||||
|
||||
Given a number of data points $N$, and a distance $\varepsilon$, DBSCAN clusters the data set by setting all data points that have $N-1$ other data points within a distance of $\varepsilon$ as *core points*. A data point that is within a distance of $\varepsilon$ from a *core point* but that does not have $N-1$ other data points within a distance of $\varepsilon$ from itself is considered an *edge point*. A cluster is then the collection of *core points* and *edge points*. Data points that have no other data points at a distance $<\varepsilon$ are considered outliers. The figure below shows a cluster with $N = 3$.
|
||||
|
||||
|
||||
|
||||
FreqAI uses `sklearn.cluster.DBSCAN` (details are available on scikit-learn's webpage [here](https://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.html) (external website)) with `min_samples` ($N$) taken as 1/4 of the no. of time points (candles) in the feature set. `eps` ($\varepsilon$) is computed automatically as the elbow point in the *k-distance graph* computed from the nearest neighbors in the pairwise distances of all data points in the feature set.
|
||||
|
||||
|
||||
### Data dimensionality reduction with Principal Component Analysis
|
||||
|
||||
You can reduce the dimensionality of your features by activating the principal_component_analysis in the config:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"feature_parameters" : {
|
||||
"principal_component_analysis": true
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
This will perform PCA on the features and reduce their dimensionality so that the explained variance of the data set is >= 0.999. Reducing data dimensionality makes training the model faster and hence allows for more up-to-date models.
|
||||
118
md/freqai-parameter-table.md
Normal file
118
md/freqai-parameter-table.md
Normal file
@ -0,0 +1,118 @@
|
||||
# Parameter table
|
||||
|
||||
The table below will list all configuration parameters available for FreqAI. Some of the parameters are exemplified in `config_examples/config_freqai.example.json`.
|
||||
|
||||
Mandatory parameters are marked as **Required** and have to be set in one of the suggested ways.
|
||||
|
||||
### General configuration parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **General configuration parameters within the `config.freqai` tree**
|
||||
| `freqai` | **Required.** <br> The parent dictionary containing all the parameters for controlling FreqAI. <br> **Datatype:** Dictionary.
|
||||
| `train_period_days` | **Required.** <br> Number of days to use for the training data (width of the sliding window). <br> **Datatype:** Positive integer.
|
||||
| `backtest_period_days` | **Required.** <br> Number of days to inference from the trained model before sliding the `train_period_days` window defined above, and retraining the model during backtesting (more info [here](freqai-running.md#backtesting)). This can be fractional days, but beware that the provided `timerange` will be divided by this number to yield the number of trainings necessary to complete the backtest. <br> **Datatype:** Float.
|
||||
| `identifier` | **Required.** <br> A unique ID for the current model. If models are saved to disk, the `identifier` allows for reloading specific pre-trained models/data. <br> **Datatype:** String.
|
||||
| `live_retrain_hours` | Frequency of retraining during dry/live runs. <br> **Datatype:** Float > 0. <br> Default: `0` (models retrain as often as possible).
|
||||
| `expiration_hours` | Avoid making predictions if a model is more than `expiration_hours` old. <br> **Datatype:** Positive integer. <br> Default: `0` (models never expire).
|
||||
| `purge_old_models` | Number of models to keep on disk (not relevant to backtesting). Default is 2, which means that dry/live runs will keep the latest 2 models on disk. Setting to 0 keeps all models. This parameter also accepts a boolean to maintain backwards compatibility. <br> **Datatype:** Integer. <br> Default: `2`.
|
||||
| `save_backtest_models` | Save models to disk when running backtesting. Backtesting operates most efficiently by saving the prediction data and reusing them directly for subsequent runs (when you wish to tune entry/exit parameters). Saving backtesting models to disk also allows to use the same model files for starting a dry/live instance with the same model `identifier`. <br> **Datatype:** Boolean. <br> Default: `False` (no models are saved).
|
||||
| `fit_live_predictions_candles` | Number of historical candles to use for computing target (label) statistics from prediction data, instead of from the training dataset (more information can be found [here](freqai-configuration.md#creating-a-dynamic-target-threshold)). <br> **Datatype:** Positive integer.
|
||||
| `continual_learning` | Use the final state of the most recently trained model as starting point for the new model, allowing for incremental learning (more information can be found [here](freqai-running.md#continual-learning)). Beware that this is currently a naive approach to incremental learning, and it has a high probability of overfitting/getting stuck in local minima while the market moves away from your model. We have the connections here primarily for experimental purposes and so that it is ready for more mature approaches to continual learning in chaotic systems like the crypto market. <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
| `write_metrics_to_disk` | Collect train timings, inference timings and cpu usage in json file. <br> **Datatype:** Boolean. <br> Default: `False`
|
||||
| `data_kitchen_thread_count` | <br> Designate the number of threads you want to use for data processing (outlier methods, normalization, etc.). This has no impact on the number of threads used for training. If user does not set it (default), FreqAI will use max number of threads - 2 (leaving 1 physical core available for Freqtrade bot and FreqUI) <br> **Datatype:** Positive integer.
|
||||
| `activate_tensorboard` | <br> Indicate whether or not to activate tensorboard for the tensorboard enabled modules (currently Reinforcment Learning, XGBoost, Catboost, and PyTorch). Tensorboard needs Torch installed, which means you will need the torch/RL docker image or you need to answer "yes" to the install question about whether or not you wish to install Torch. <br> **Datatype:** Boolean. <br> Default: `True`.
|
||||
| `wait_for_training_iteration_on_reload` | <br> When using /reload or ctrl-c, wait for the current training iteration to finish before completing graceful shutdown. If set to `False`, FreqAI will break the current training iteration, allowing you to shutdown gracefully more quickly, but you will lose your current training iteration. <br> **Datatype:** Boolean. <br> Default: `True`.
|
||||
|
||||
### Feature parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Feature parameters within the `freqai.feature_parameters` sub dictionary**
|
||||
| `feature_parameters` | A dictionary containing the parameters used to engineer the feature set. Details and examples are shown [here](freqai-feature-engineering.md). <br> **Datatype:** Dictionary.
|
||||
| `include_timeframes` | A list of timeframes that all indicators in `feature_engineering_expand_*()` will be created for. The list is added as features to the base indicators dataset. <br> **Datatype:** List of timeframes (strings).
|
||||
| `include_corr_pairlist` | A list of correlated coins that FreqAI will add as additional features to all `pair_whitelist` coins. All indicators set in `feature_engineering_expand_*()` during feature engineering (see details [here](freqai-feature-engineering.md)) will be created for each correlated coin. The correlated coins features are added to the base indicators dataset. <br> **Datatype:** List of assets (strings).
|
||||
| `label_period_candles` | Number of candles into the future that the labels are created for. This can be used in `set_freqai_targets()` (see `templates/FreqaiExampleStrategy.py` for detailed usage). This parameter is not necessarily required, you can create custom labels and choose whether to make use of this parameter or not. Please see `templates/FreqaiExampleStrategy.py` to see the example usage. <br> **Datatype:** Positive integer.
|
||||
| `include_shifted_candles` | Add features from previous candles to subsequent candles with the intent of adding historical information. If used, FreqAI will duplicate and shift all features from the `include_shifted_candles` previous candles so that the information is available for the subsequent candle. <br> **Datatype:** Positive integer.
|
||||
| `weight_factor` | Weight training data points according to their recency (see details [here](freqai-feature-engineering.md#weighting-features-for-temporal-importance)). <br> **Datatype:** Positive float (typically < 1).
|
||||
| `indicator_max_period_candles` | **No longer used (#7325)**. Replaced by `startup_candle_count` which is set in the [strategy](freqai-configuration.md#building-a-freqai-strategy). `startup_candle_count` is timeframe independent and defines the maximum *period* used in `feature_engineering_*()` for indicator creation. FreqAI uses this parameter together with the maximum timeframe in `include_time_frames` to calculate how many data points to download such that the first data point does not include a NaN. <br> **Datatype:** Positive integer.
|
||||
| `indicator_periods_candles` | Time periods to calculate indicators for. The indicators are added to the base indicator dataset. <br> **Datatype:** List of positive integers.
|
||||
| `principal_component_analysis` | Automatically reduce the dimensionality of the data set using Principal Component Analysis. See details about how it works [here](freqai-feature-engineering.md#data-dimensionality-reduction-with-principal-component-analysis) <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
| `plot_feature_importances` | Create a feature importance plot for each model for the top/bottom `plot_feature_importances` number of features. Plot is stored in `user_data/models/<identifier>/sub-train-<COIN>_<timestamp>.html`. <br> **Datatype:** Integer. <br> Default: `0`.
|
||||
| `DI_threshold` | Activates the use of the Dissimilarity Index for outlier detection when set to > 0. See details about how it works [here](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di). <br> **Datatype:** Positive float (typically < 1).
|
||||
| `use_SVM_to_remove_outliers` | Train a support vector machine to detect and remove outliers from the training dataset, as well as from incoming data points. See details about how it works [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm). <br> **Datatype:** Boolean.
|
||||
| `svm_params` | All parameters available in Sklearn's `SGDOneClassSVM()`. See details about some select parameters [here](freqai-feature-engineering.md#identifying-outliers-using-a-support-vector-machine-svm). <br> **Datatype:** Dictionary.
|
||||
| `use_DBSCAN_to_remove_outliers` | Cluster data using the DBSCAN algorithm to identify and remove outliers from training and prediction data. See details about how it works [here](freqai-feature-engineering.md#identifying-outliers-with-dbscan). <br> **Datatype:** Boolean.
|
||||
| `noise_standard_deviation` | If set, FreqAI adds noise to the training features with the aim of preventing overfitting. FreqAI generates random deviates from a gaussian distribution with a standard deviation of `noise_standard_deviation` and adds them to all data points. `noise_standard_deviation` should be kept relative to the normalized space, i.e., between -1 and 1. In other words, since data in FreqAI is always normalized to be between -1 and 1, `noise_standard_deviation: 0.05` would result in 32% of the data being randomly increased/decreased by more than 2.5% (i.e., the percent of data falling within the first standard deviation). <br> **Datatype:** Integer. <br> Default: `0`.
|
||||
| `outlier_protection_percentage` | Enable to prevent outlier detection methods from discarding too much data. If more than `outlier_protection_percentage` % of points are detected as outliers by the SVM or DBSCAN, FreqAI will log a warning message and ignore outlier detection, i.e., the original dataset will be kept intact. If the outlier protection is triggered, no predictions will be made based on the training dataset. <br> **Datatype:** Float. <br> Default: `30`.
|
||||
| `reverse_train_test_order` | Split the feature dataset (see below) and use the latest data split for training and test on historical split of the data. This allows the model to be trained up to the most recent data point, while avoiding overfitting. However, you should be careful to understand the unorthodox nature of this parameter before employing it. <br> **Datatype:** Boolean. <br> Default: `False` (no reversal).
|
||||
| `shuffle_after_split` | Split the data into train and test sets, and then shuffle both sets individually. <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
| `buffer_train_data_candles` | Cut `buffer_train_data_candles` off the beginning and end of the training data *after* the indicators were populated. The main example use is when predicting maxima and minima, the argrelextrema function cannot know the maxima/minima at the edges of the timerange. To improve model accuracy, it is best to compute argrelextrema on the full timerange and then use this function to cut off the edges (buffer) by the kernel. In another case, if the targets are set to a shifted price movement, this buffer is unnecessary because the shifted candles at the end of the timerange will be NaN and FreqAI will automatically cut those off of the training dataset.<br> **Datatype:** Integer. <br> Default: `0`.
|
||||
|
||||
### Data split parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Data split parameters within the `freqai.data_split_parameters` sub dictionary**
|
||||
| `data_split_parameters` | Include any additional parameters available from scikit-learn `test_train_split()`, which are shown [here](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html) (external website). <br> **Datatype:** Dictionary.
|
||||
| `test_size` | The fraction of data that should be used for testing instead of training. <br> **Datatype:** Positive float < 1.
|
||||
| `shuffle` | Shuffle the training data points during training. Typically, to not remove the chronological order of data in time-series forecasting, this is set to `False`. <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
|
||||
### Model training parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Model training parameters within the `freqai.model_training_parameters` sub dictionary**
|
||||
| `model_training_parameters` | A flexible dictionary that includes all parameters available by the selected model library. For example, if you use `LightGBMRegressor`, this dictionary can contain any parameter available by the `LightGBMRegressor` [here](https://lightgbm.readthedocs.io/en/latest/pythonapi/lightgbm.LGBMRegressor.html) (external website). If you select a different model, this dictionary can contain any parameter from that model. A list of the currently available models can be found [here](freqai-configuration.md#using-different-prediction-models). <br> **Datatype:** Dictionary.
|
||||
| `n_estimators` | The number of boosted trees to fit in the training of the model. <br> **Datatype:** Integer.
|
||||
| `learning_rate` | Boosting learning rate during training of the model. <br> **Datatype:** Float.
|
||||
| `n_jobs`, `thread_count`, `task_type` | Set the number of threads for parallel processing and the `task_type` (`gpu` or `cpu`). Different model libraries use different parameter names. <br> **Datatype:** Float.
|
||||
|
||||
### Reinforcement Learning parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Reinforcement Learning Parameters within the `freqai.rl_config` sub dictionary**
|
||||
| `rl_config` | A dictionary containing the control parameters for a Reinforcement Learning model. <br> **Datatype:** Dictionary.
|
||||
| `train_cycles` | Training time steps will be set based on the `train_cycles * number of training data points. <br> **Datatype:** Integer.
|
||||
| `max_trade_duration_candles`| Guides the agent training to keep trades below desired length. Example usage shown in `prediction_models/ReinforcementLearner.py` within the customizable `calculate_reward()` function. <br> **Datatype:** int.
|
||||
| `model_type` | Model string from stable_baselines3 or SBcontrib. Available strings include: `'TRPO', 'ARS', 'RecurrentPPO', 'MaskablePPO', 'PPO', 'A2C', 'DQN'`. User should ensure that `model_training_parameters` match those available to the corresponding stable_baselines3 model by visiting their documentation. [PPO doc](https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html) (external website) <br> **Datatype:** string.
|
||||
| `policy_type` | One of the available policy types from stable_baselines3 <br> **Datatype:** string.
|
||||
| `max_training_drawdown_pct` | The maximum drawdown that the agent is allowed to experience during training. <br> **Datatype:** float. <br> Default: 0.8
|
||||
| `cpu_count` | Number of threads/cpus to dedicate to the Reinforcement Learning training process (depending on if `ReinforcementLearning_multiproc` is selected or not). Recommended to leave this untouched, by default, this value is set to the total number of physical cores minus 1. <br> **Datatype:** int.
|
||||
| `model_reward_parameters` | Parameters used inside the customizable `calculate_reward()` function in `ReinforcementLearner.py` <br> **Datatype:** int.
|
||||
| `add_state_info` | Tell FreqAI to include state information in the feature set for training and inferencing. The current state variables include trade duration, current profit, trade position. This is only available in dry/live runs, and is automatically switched to false for backtesting. <br> **Datatype:** bool. <br> Default: `False`.
|
||||
| `net_arch` | Network architecture which is well described in [`stable_baselines3` doc](https://stable-baselines3.readthedocs.io/en/master/guide/custom_policy.html#examples). In summary: `[<shared layers>, dict(vf=[<non-shared value network layers>], pi=[<non-shared policy network layers>])]`. By default this is set to `[128, 128]`, which defines 2 shared hidden layers with 128 units each.
|
||||
| `randomize_starting_position` | Randomize the starting point of each episode to avoid overfitting. <br> **Datatype:** bool. <br> Default: `False`.
|
||||
| `drop_ohlc_from_features` | Do not include the normalized ohlc data in the feature set passed to the agent during training (ohlc will still be used for driving the environment in all cases) <br> **Datatype:** Boolean. <br> **Default:** `False`
|
||||
| `progress_bar` | Display a progress bar with the current progress, elapsed time and estimated remaining time. <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
|
||||
### PyTorch parameters
|
||||
|
||||
#### general
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Model training parameters within the `freqai.model_training_parameters` sub dictionary**
|
||||
| `learning_rate` | Learning rate to be passed to the optimizer. <br> **Datatype:** float. <br> Default: `3e-4`.
|
||||
| `model_kwargs` | Parameters to be passed to the model class. <br> **Datatype:** dict. <br> Default: `{}`.
|
||||
| `trainer_kwargs` | Parameters to be passed to the trainer class. <br> **Datatype:** dict. <br> Default: `{}`.
|
||||
|
||||
#### trainer_kwargs
|
||||
|
||||
| Parameter | Description |
|
||||
|--------------|-------------|
|
||||
| | **Model training parameters within the `freqai.model_training_parameters.model_kwargs` sub dictionary**
|
||||
| `n_epochs` | The `n_epochs` parameter is a crucial setting in the PyTorch training loop that determines the number of times the entire training dataset will be used to update the model's parameters. An epoch represents one full pass through the entire training dataset. Overrides `n_steps`. Either `n_epochs` or `n_steps` must be set. <br><br> **Datatype:** int. optional. <br> Default: `10`.
|
||||
| `n_steps` | An alternative way of setting `n_epochs` - the number of training iterations to run. Iteration here refer to the number of times we call `optimizer.step()`. Ignored if `n_epochs` is set. A simplified version of the function: <br><br> n_epochs = n_steps / (n_obs / batch_size) <br><br> The motivation here is that `n_steps` is easier to optimize and keep stable across different n_obs - the number of data points. <br> <br> **Datatype:** int. optional. <br> Default: `None`.
|
||||
| `batch_size` | The size of the batches to use during training. <br><br> **Datatype:** int. <br> Default: `64`.
|
||||
|
||||
|
||||
### Additional parameters
|
||||
|
||||
| Parameter | Description |
|
||||
|------------|-------------|
|
||||
| | **Extraneous parameters**
|
||||
| `freqai.keras` | If the selected model makes use of Keras (typical for TensorFlow-based prediction models), this flag needs to be activated so that the model save/loading follows Keras standards. <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
| `freqai.conv_width` | The width of a neural network input tensor. This replaces the need for shifting candles (`include_shifted_candles`) by feeding in historical data points as the second dimension of the tensor. Technically, this parameter can also be used for regressors, but it only adds computational overhead and does not change the model training/prediction. <br> **Datatype:** Integer. <br> Default: `2`.
|
||||
| `freqai.reduce_df_footprint` | Recast all numeric columns to float32/int32, with the objective of reducing ram/disk usage and decreasing train/inference timing. This parameter is set in the main level of the Freqtrade configuration file (not inside FreqAI). <br> **Datatype:** Boolean. <br> Default: `False`.
|
||||
280
md/freqai-reinforcement-learning.md
Normal file
280
md/freqai-reinforcement-learning.md
Normal file
@ -0,0 +1,280 @@
|
||||
# Reinforcement Learning
|
||||
|
||||
!!! Note "Installation size"
|
||||
Reinforcement learning dependencies include large packages such as `torch`, which should be explicitly requested during `./setup.sh -i` by answering "y" to the question "Do you also want dependencies for freqai-rl (~700mb additional space required) [y/N]?".
|
||||
Users who prefer docker should ensure they use the docker image appended with `_freqairl`.
|
||||
|
||||
## Background and terminology
|
||||
|
||||
### What is RL and why does FreqAI need it?
|
||||
|
||||
Reinforcement learning involves two important components, the *agent* and the training *environment*. During agent training, the agent moves through historical data candle by candle, always making 1 of a set of actions: Long entry, long exit, short entry, short exit, neutral). During this training process, the environment tracks the performance of these actions and rewards the agent according to a custom user made `calculate_reward()` (here we offer a default reward for users to build on if they wish [details here](#creating-a-custom-reward-function)). The reward is used to train weights in a neural network.
|
||||
|
||||
A second important component of the FreqAI RL implementation is the use of *state* information. State information is fed into the network at each step, including current profit, current position, and current trade duration. These are used to train the agent in the training environment, and to reinforce the agent in dry/live (this functionality is not available in backtesting). *FreqAI + Freqtrade is a perfect match for this reinforcing mechanism since this information is readily available in live deployments.*
|
||||
|
||||
Reinforcement learning is a natural progression for FreqAI, since it adds a new layer of adaptivity and market reactivity that Classifiers and Regressors cannot match. However, Classifiers and Regressors have strengths that RL does not have such as robust predictions. Improperly trained RL agents may find "cheats" and "tricks" to maximize reward without actually winning any trades. For this reason, RL is more complex and demands a higher level of understanding than typical Classifiers and Regressors.
|
||||
|
||||
### The RL interface
|
||||
|
||||
With the current framework, we aim to expose the training environment via the common "prediction model" file, which is a user inherited `BaseReinforcementLearner` object (e.g. `freqai/prediction_models/ReinforcementLearner`). Inside this user class, the RL environment is available and customized via `MyRLEnv` as [shown below](#creating-a-custom-reward-function).
|
||||
|
||||
We envision the majority of users focusing their effort on creative design of the `calculate_reward()` function [details here](#creating-a-custom-reward-function), while leaving the rest of the environment untouched. Other users may not touch the environment at all, and they will only play with the configuration settings and the powerful feature engineering that already exists in FreqAI. Meanwhile, we enable advanced users to create their own model classes entirely.
|
||||
|
||||
The framework is built on stable_baselines3 (torch) and OpenAI gym for the base environment class. But generally speaking, the model class is well isolated. Thus, the addition of competing libraries can be easily integrated into the existing framework. For the environment, it is inheriting from `gym.Env` which means that it is necessary to write an entirely new environment in order to switch to a different library.
|
||||
|
||||
### Important considerations
|
||||
|
||||
As explained above, the agent is "trained" in an artificial trading "environment". In our case, that environment may seem quite similar to a real Freqtrade backtesting environment, but it is *NOT*. In fact, the RL training environment is much more simplified. It does not incorporate any of the complicated strategy logic, such as callbacks like `custom_exit`, `custom_stoploss`, leverage controls, etc. The RL environment is instead a very "raw" representation of the true market, where the agent has free will to learn the policy (read: stoploss, take profit, etc.) which is enforced by the `calculate_reward()`. Thus, it is important to consider that the agent training environment is not identical to the real world.
|
||||
|
||||
## Running Reinforcement Learning
|
||||
|
||||
Setting up and running a Reinforcement Learning model is the same as running a Regressor or Classifier. The same two flags, `--freqaimodel` and `--strategy`, must be defined on the command line:
|
||||
|
||||
```bash
|
||||
freqtrade trade --freqaimodel ReinforcementLearner --strategy MyRLStrategy --config config.json
|
||||
```
|
||||
|
||||
where `ReinforcementLearner` will use the templated `ReinforcementLearner` from `freqai/prediction_models/ReinforcementLearner` (or a custom user defined one located in `user_data/freqaimodels`). The strategy, on the other hand, follows the same base [feature engineering](freqai-feature-engineering.md) with `feature_engineering_*` as a typical Regressor. The difference lies in the creation of the targets, Reinforcement Learning doesn't require them. However, FreqAI requires a default (neutral) value to be set in the action column:
|
||||
|
||||
```python
|
||||
def set_freqai_targets(self, dataframe, **kwargs) -> DataFrame:
|
||||
"""
|
||||
*Only functional with FreqAI enabled strategies*
|
||||
Required function to set the targets for the model.
|
||||
All targets must be prepended with `&` to be recognized by the FreqAI internals.
|
||||
|
||||
More details about feature engineering available:
|
||||
|
||||
https://www.freqtrade.io/en/latest/freqai-feature-engineering
|
||||
|
||||
:param df: strategy dataframe which will receive the targets
|
||||
usage example: dataframe["&-target"] = dataframe["close"].shift(-1) / dataframe["close"]
|
||||
"""
|
||||
# For RL, there are no direct targets to set. This is filler (neutral)
|
||||
# until the agent sends an action.
|
||||
dataframe["&-action"] = 0
|
||||
return dataframe
|
||||
```
|
||||
|
||||
Most of the function remains the same as for typical Regressors, however, the function below shows how the strategy must pass the raw price data to the agent so that it has access to raw OHLCV in the training environment:
|
||||
|
||||
```python
|
||||
def feature_engineering_standard(self, dataframe: DataFrame, **kwargs) -> DataFrame:
|
||||
# The following features are necessary for RL models
|
||||
dataframe[f"%-raw_close"] = dataframe["close"]
|
||||
dataframe[f"%-raw_open"] = dataframe["open"]
|
||||
dataframe[f"%-raw_high"] = dataframe["high"]
|
||||
dataframe[f"%-raw_low"] = dataframe["low"]
|
||||
return dataframe
|
||||
```
|
||||
|
||||
Finally, there is no explicit "label" to make - instead it is necessary to assign the `&-action` column which will contain the agent's actions when accessed in `populate_entry/exit_trends()`. In the present example, the neutral action to 0. This value should align with the environment used. FreqAI provides two environments, both use 0 as the neutral action.
|
||||
|
||||
After users realize there are no labels to set, they will soon understand that the agent is making its "own" entry and exit decisions. This makes strategy construction rather simple. The entry and exit signals come from the agent in the form of an integer - which are used directly to decide entries and exits in the strategy:
|
||||
|
||||
```python
|
||||
def populate_entry_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
|
||||
|
||||
enter_long_conditions = [df["do_predict"] == 1, df["&-action"] == 1]
|
||||
|
||||
if enter_long_conditions:
|
||||
df.loc[
|
||||
reduce(lambda x, y: x & y, enter_long_conditions), ["enter_long", "enter_tag"]
|
||||
] = (1, "long")
|
||||
|
||||
enter_short_conditions = [df["do_predict"] == 1, df["&-action"] == 3]
|
||||
|
||||
if enter_short_conditions:
|
||||
df.loc[
|
||||
reduce(lambda x, y: x & y, enter_short_conditions), ["enter_short", "enter_tag"]
|
||||
] = (1, "short")
|
||||
|
||||
return df
|
||||
|
||||
def populate_exit_trend(self, df: DataFrame, metadata: dict) -> DataFrame:
|
||||
exit_long_conditions = [df["do_predict"] == 1, df["&-action"] == 2]
|
||||
if exit_long_conditions:
|
||||
df.loc[reduce(lambda x, y: x & y, exit_long_conditions), "exit_long"] = 1
|
||||
|
||||
exit_short_conditions = [df["do_predict"] == 1, df["&-action"] == 4]
|
||||
if exit_short_conditions:
|
||||
df.loc[reduce(lambda x, y: x & y, exit_short_conditions), "exit_short"] = 1
|
||||
|
||||
return df
|
||||
```
|
||||
|
||||
It is important to consider that `&-action` depends on which environment they choose to use. The example above shows 5 actions, where 0 is neutral, 1 is enter long, 2 is exit long, 3 is enter short and 4 is exit short.
|
||||
|
||||
## Configuring the Reinforcement Learner
|
||||
|
||||
In order to configure the `Reinforcement Learner` the following dictionary must exist in the `freqai` config:
|
||||
|
||||
```json
|
||||
"rl_config": {
|
||||
"train_cycles": 25,
|
||||
"add_state_info": true,
|
||||
"max_trade_duration_candles": 300,
|
||||
"max_training_drawdown_pct": 0.02,
|
||||
"cpu_count": 8,
|
||||
"model_type": "PPO",
|
||||
"policy_type": "MlpPolicy",
|
||||
"model_reward_parameters": {
|
||||
"rr": 1,
|
||||
"profit_aim": 0.025
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Parameter details can be found [here](freqai-parameter-table.md), but in general the `train_cycles` decides how many times the agent should cycle through the candle data in its artificial environment to train weights in the model. `model_type` is a string which selects one of the available models in [stable_baselines](https://stable-baselines3.readthedocs.io/en/master/)(external link).
|
||||
|
||||
!!! Note
|
||||
If you would like to experiment with `continual_learning`, then you should set that value to `true` in the main `freqai` configuration dictionary. This will tell the Reinforcement Learning library to continue training new models from the final state of previous models, instead of retraining new models from scratch each time a retrain is initiated.
|
||||
|
||||
!!! Note
|
||||
Remember that the general `model_training_parameters` dictionary should contain all the model hyperparameter customizations for the particular `model_type`. For example, `PPO` parameters can be found [here](https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html).
|
||||
|
||||
## Creating a custom reward function
|
||||
|
||||
!!! danger "Not for production"
|
||||
Warning!
|
||||
The reward function provided with the Freqtrade source code is a showcase of functionality designed to show/test as many possible environment control features as possible. It is also designed to run quickly on small computers. This is a benchmark, it is *not* for live production. Please beware that you will need to create your own custom_reward() function or use a template built by other users outside of the Freqtrade source code.
|
||||
|
||||
As you begin to modify the strategy and the prediction model, you will quickly realize some important differences between the Reinforcement Learner and the Regressors/Classifiers. Firstly, the strategy does not set a target value (no labels!). Instead, you set the `calculate_reward()` function inside the `MyRLEnv` class (see below). A default `calculate_reward()` is provided inside `prediction_models/ReinforcementLearner.py` to demonstrate the necessary building blocks for creating rewards, but this is *not* designed for production. Users *must* create their own custom reinforcement learning model class or use a pre-built one from outside the Freqtrade source code and save it to `user_data/freqaimodels`. It is inside the `calculate_reward()` where creative theories about the market can be expressed. For example, you can reward your agent when it makes a winning trade, and penalize the agent when it makes a losing trade. Or perhaps, you wish to reward the agent for entering trades, and penalize the agent for sitting in trades too long. Below we show examples of how these rewards are all calculated:
|
||||
|
||||
!!! note "Hint"
|
||||
The best reward functions are ones that are continuously differentiable, and well scaled. In other words, adding a single large negative penalty to a rare event is not a good idea, and the neural net will not be able to learn that function. Instead, it is better to add a small negative penalty to a common event. This will help the agent learn faster. Not only this, but you can help improve the continuity of your rewards/penalties by having them scale with severity according to some linear/exponential functions. In other words, you'd slowly scale the penalty as the duration of the trade increases. This is better than a single large penalty occurring at a single point in time.
|
||||
|
||||
```python
|
||||
from freqtrade.freqai.prediction_models.ReinforcementLearner import ReinforcementLearner
|
||||
from freqtrade.freqai.RL.Base5ActionRLEnv import Actions, Base5ActionRLEnv, Positions
|
||||
|
||||
|
||||
class MyCoolRLModel(ReinforcementLearner):
|
||||
"""
|
||||
User created RL prediction model.
|
||||
|
||||
Save this file to `freqtrade/user_data/freqaimodels`
|
||||
|
||||
then use it with:
|
||||
|
||||
freqtrade trade --freqaimodel MyCoolRLModel --config config.json --strategy SomeCoolStrat
|
||||
|
||||
Here the users can override any of the functions
|
||||
available in the `IFreqaiModel` inheritance tree. Most importantly for RL, this
|
||||
is where the user overrides `MyRLEnv` (see below), to define custom
|
||||
`calculate_reward()` function, or to override any other parts of the environment.
|
||||
|
||||
This class also allows users to override any other part of the IFreqaiModel tree.
|
||||
For example, the user can override `def fit()` or `def train()` or `def predict()`
|
||||
to take fine-tuned control over these processes.
|
||||
|
||||
Another common override may be `def data_cleaning_predict()` where the user can
|
||||
take fine-tuned control over the data handling pipeline.
|
||||
"""
|
||||
class MyRLEnv(Base5ActionRLEnv):
|
||||
"""
|
||||
User made custom environment. This class inherits from BaseEnvironment and gym.Env.
|
||||
Users can override any functions from those parent classes. Here is an example
|
||||
of a user customized `calculate_reward()` function.
|
||||
|
||||
Warning!
|
||||
This is function is a showcase of functionality designed to show as many possible
|
||||
environment control features as possible. It is also designed to run quickly
|
||||
on small computers. This is a benchmark, it is *not* for live production.
|
||||
"""
|
||||
def calculate_reward(self, action: int) -> float:
|
||||
# first, penalize if the action is not valid
|
||||
if not self._is_valid(action):
|
||||
return -2
|
||||
pnl = self.get_unrealized_profit()
|
||||
|
||||
factor = 100
|
||||
|
||||
pair = self.pair.replace(':', '')
|
||||
|
||||
# you can use feature values from dataframe
|
||||
# Assumes the shifted RSI indicator has been generated in the strategy.
|
||||
rsi_now = self.raw_features[f"%-rsi-period_10_shift-1_{pair}_"
|
||||
f"{self.config['timeframe']}"].iloc[self._current_tick]
|
||||
|
||||
# reward agent for entering trades
|
||||
if (action in (Actions.Long_enter.value, Actions.Short_enter.value)
|
||||
and self._position == Positions.Neutral):
|
||||
if rsi_now < 40:
|
||||
factor = 40 / rsi_now
|
||||
else:
|
||||
factor = 1
|
||||
return 25 * factor
|
||||
|
||||
# discourage agent from not entering trades
|
||||
if action == Actions.Neutral.value and self._position == Positions.Neutral:
|
||||
return -1
|
||||
max_trade_duration = self.rl_config.get('max_trade_duration_candles', 300)
|
||||
trade_duration = self._current_tick - self._last_trade_tick
|
||||
if trade_duration <= max_trade_duration:
|
||||
factor *= 1.5
|
||||
elif trade_duration > max_trade_duration:
|
||||
factor *= 0.5
|
||||
# discourage sitting in position
|
||||
if self._position in (Positions.Short, Positions.Long) and \
|
||||
action == Actions.Neutral.value:
|
||||
return -1 * trade_duration / max_trade_duration
|
||||
# close long
|
||||
if action == Actions.Long_exit.value and self._position == Positions.Long:
|
||||
if pnl > self.profit_aim * self.rr:
|
||||
factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
|
||||
return float(pnl * factor)
|
||||
# close short
|
||||
if action == Actions.Short_exit.value and self._position == Positions.Short:
|
||||
if pnl > self.profit_aim * self.rr:
|
||||
factor *= self.rl_config['model_reward_parameters'].get('win_reward_factor', 2)
|
||||
return float(pnl * factor)
|
||||
return 0.
|
||||
```
|
||||
|
||||
## Using Tensorboard
|
||||
|
||||
Reinforcement Learning models benefit from tracking training metrics. FreqAI has integrated Tensorboard to allow users to track training and evaluation performance across all coins and across all retrainings. Tensorboard is activated via the following command:
|
||||
|
||||
```bash
|
||||
tensorboard --logdir user_data/models/unique-id
|
||||
```
|
||||
|
||||
where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell to view the output in the browser at 127.0.0.1:6006 (6006 is the default port used by Tensorboard).
|
||||
|
||||
|
||||
|
||||
## Custom logging
|
||||
|
||||
FreqAI also provides a built in episodic summary logger called `self.tensorboard_log` for adding custom information to the Tensorboard log. By default, this function is already called once per step inside the environment to record the agent actions. All values accumulated for all steps in a single episode are reported at the conclusion of each episode, followed by a full reset of all metrics to 0 in preparation for the subsequent episode.
|
||||
|
||||
`self.tensorboard_log` can also be used anywhere inside the environment, for example, it can be added to the `calculate_reward` function to collect more detailed information about how often various parts of the reward were called:
|
||||
|
||||
```python
|
||||
class MyRLEnv(Base5ActionRLEnv):
|
||||
"""
|
||||
User made custom environment. This class inherits from BaseEnvironment and gym.Env.
|
||||
Users can override any functions from those parent classes. Here is an example
|
||||
of a user customized `calculate_reward()` function.
|
||||
"""
|
||||
def calculate_reward(self, action: int) -> float:
|
||||
if not self._is_valid(action):
|
||||
self.tensorboard_log("invalid")
|
||||
return -2
|
||||
|
||||
```
|
||||
|
||||
!!! Note
|
||||
The `self.tensorboard_log()` function is designed for tracking incremented objects only i.e. events, actions inside the training environment. If the event of interest is a float, the float can be passed as the second argument e.g. `self.tensorboard_log("float_metric1", 0.23)`. In this case the metric values are not incremented.
|
||||
|
||||
## Choosing a base environment
|
||||
|
||||
FreqAI provides three base environments, `Base3ActionRLEnvironment`, `Base4ActionEnvironment` and `Base5ActionEnvironment`. As the names imply, the environments are customized for agents that can select from 3, 4 or 5 actions. The `Base3ActionEnvironment` is the simplest, the agent can select from hold, long, or short. This environment can also be used for long-only bots (it automatically follows the `can_short` flag from the strategy), where long is the enter condition and short is the exit condition. Meanwhile, in the `Base4ActionEnvironment`, the agent can enter long, enter short, hold neutral, or exit position. Finally, in the `Base5ActionEnvironment`, the agent has the same actions as Base4, but instead of a single exit action, it separates exit long and exit short. The main changes stemming from the environment selection include:
|
||||
|
||||
* the actions available in the `calculate_reward`
|
||||
* the actions consumed by the user strategy
|
||||
|
||||
All of the FreqAI provided environments inherit from an action/position agnostic environment object called the `BaseEnvironment`, which contains all shared logic. The architecture is designed to be easily customized. The simplest customization is the `calculate_reward()` (see details [here](#creating-a-custom-reward-function)). However, the customizations can be further extended into any of the functions inside the environment. You can do this by simply overriding those functions inside your `MyRLEnv` in the prediction model file. Or for more advanced customizations, it is encouraged to create an entirely new environment inherited from `BaseEnvironment`.
|
||||
|
||||
!!! Note
|
||||
Only the `Base3ActionRLEnv` can do long-only training/trading (set the user strategy attribute `can_short = False`).
|
||||
187
md/freqai-running.md
Normal file
187
md/freqai-running.md
Normal file
@ -0,0 +1,187 @@
|
||||
# Running FreqAI
|
||||
|
||||
There are two ways to train and deploy an adaptive machine learning model - live deployment and historical backtesting. In both cases, FreqAI runs/simulates periodic retraining of models as shown in the following figure:
|
||||
|
||||
|
||||
|
||||
## Live deployments
|
||||
|
||||
FreqAI can be run dry/live using the following command:
|
||||
|
||||
```bash
|
||||
freqtrade trade --strategy FreqaiExampleStrategy --config config_freqai.example.json --freqaimodel LightGBMRegressor
|
||||
```
|
||||
|
||||
When launched, FreqAI will start training a new model, with a new `identifier`, based on the config settings. Following training, the model will be used to make predictions on incoming candles until a new model is available. New models are typically generated as often as possible, with FreqAI managing an internal queue of the coin pairs to try to keep all models equally up to date. FreqAI will always use the most recently trained model to make predictions on incoming live data. If you do not want FreqAI to retrain new models as often as possible, you can set `live_retrain_hours` to tell FreqAI to wait at least that number of hours before training a new model. Additionally, you can set `expired_hours` to tell FreqAI to avoid making predictions on models that are older than that number of hours.
|
||||
|
||||
Trained models are by default saved to disk to allow for reuse during backtesting or after a crash. You can opt to [purge old models](#purging-old-model-data) to save disk space by setting `"purge_old_models": true` in the config.
|
||||
|
||||
To start a dry/live run from a saved backtest model (or from a previously crashed dry/live session), you only need to specify the `identifier` of the specific model:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"identifier": "example",
|
||||
"live_retrain_hours": 0.5
|
||||
}
|
||||
```
|
||||
|
||||
In this case, although FreqAI will initiate with a pre-trained model, it will still check to see how much time has elapsed since the model was trained. If a full `live_retrain_hours` has elapsed since the end of the loaded model, FreqAI will start training a new model.
|
||||
|
||||
### Automatic data download
|
||||
|
||||
FreqAI automatically downloads the proper amount of data needed to ensure training of a model through the defined `train_period_days` and `startup_candle_count` (see the [parameter table](freqai-parameter-table.md) for detailed descriptions of these parameters).
|
||||
|
||||
### Saving prediction data
|
||||
|
||||
All predictions made during the lifetime of a specific `identifier` model are stored in `historic_predictions.pkl` to allow for reloading after a crash or changes made to the config.
|
||||
|
||||
### Purging old model data
|
||||
|
||||
FreqAI stores new model files after each successful training. These files become obsolete as new models are generated to adapt to new market conditions. If you are planning to leave FreqAI running for extended periods of time with high frequency retraining, you should enable `purge_old_models` in the config:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"purge_old_models": 4,
|
||||
}
|
||||
```
|
||||
|
||||
This will automatically purge all models older than the four most recently trained ones to save disk space. Inputing "0" will never purge any models.
|
||||
|
||||
## Backtesting
|
||||
|
||||
The FreqAI backtesting module can be executed with the following command:
|
||||
|
||||
```bash
|
||||
freqtrade backtesting --strategy FreqaiExampleStrategy --strategy-path freqtrade/templates --config config_examples/config_freqai.example.json --freqaimodel LightGBMRegressor --timerange 20210501-20210701
|
||||
```
|
||||
|
||||
If this command has never been executed with the existing config file, FreqAI will train a new model
|
||||
for each pair, for each backtesting window within the expanded `--timerange`.
|
||||
|
||||
Backtesting mode requires [downloading the necessary data](#downloading-data-to-cover-the-full-backtest-period) before deployment (unlike in dry/live mode where FreqAI handles the data downloading automatically). You should be careful to consider that the time range of the downloaded data is more than the backtesting time range. This is because FreqAI needs data prior to the desired backtesting time range in order to train a model to be ready to make predictions on the first candle of the set backtesting time range. More details on how to calculate the data to download can be found [here](#deciding-the-size-of-the-sliding-training-window-and-backtesting-duration).
|
||||
|
||||
!!! Note "Model reuse"
|
||||
Once the training is completed, you can execute the backtesting again with the same config file and
|
||||
FreqAI will find the trained models and load them instead of spending time training. This is useful
|
||||
if you want to tweak (or even hyperopt) buy and sell criteria inside the strategy. If you
|
||||
*want* to retrain a new model with the same config file, you should simply change the `identifier`.
|
||||
This way, you can return to using any model you wish by simply specifying the `identifier`.
|
||||
|
||||
!!! Note
|
||||
Backtesting calls `set_freqai_targets()` one time for each backtest window (where the number of windows is the full backtest timerange divided by the `backtest_period_days` parameter). Doing this means that the targets simulate dry/live behavior without look ahead bias. However, the definition of the features in `feature_engineering_*()` is performed once on the entire training timerange. This means that you should be sure that features do not look-ahead into the future.
|
||||
More details about look-ahead bias can be found in [Common Mistakes](strategy-customization.md#common-mistakes-when-developing-strategies).
|
||||
|
||||
---
|
||||
|
||||
### Saving backtesting prediction data
|
||||
|
||||
To allow for tweaking your strategy (**not** the features!), FreqAI will automatically save the predictions during backtesting so that they can be reused for future backtests and live runs using the same `identifier` model. This provides a performance enhancement geared towards enabling **high-level hyperopting** of entry/exit criteria.
|
||||
|
||||
An additional directory called `backtesting_predictions`, which contains all the predictions stored in `feather` format, will be created in the `unique-id` folder.
|
||||
|
||||
To change your **features**, you **must** set a new `identifier` in the config to signal to FreqAI to train new models.
|
||||
|
||||
To save the models generated during a particular backtest so that you can start a live deployment from one of them instead of training a new model, you must set `save_backtest_models` to `True` in the config.
|
||||
|
||||
!!! Note
|
||||
To ensure that the model can be reused, freqAI will call your strategy with a dataframe of length 1.
|
||||
If your strategy requires more data than this to generate the same features, you can't reuse backtest predictions for live deployment and need to update your `identifier` for each new backtest.
|
||||
|
||||
### Backtest live collected predictions
|
||||
|
||||
FreqAI allow you to reuse live historic predictions through the backtest parameter `--freqai-backtest-live-models`. This can be useful when you want to reuse predictions generated in dry/run for comparison or other study.
|
||||
|
||||
The `--timerange` parameter must not be informed, as it will be automatically calculated through the data in the historic predictions file.
|
||||
|
||||
### Downloading data to cover the full backtest period
|
||||
|
||||
For live/dry deployments, FreqAI will download the necessary data automatically. However, to use backtesting functionality, you need to download the necessary data using `download-data` (details [here](data-download.md#data-downloading)). You need to pay careful attention to understanding how much *additional* data needs to be downloaded to ensure that there is a sufficient amount of training data *before* the start of the backtesting time range. The amount of additional data can be roughly estimated by moving the start date of the time range backwards by `train_period_days` and the `startup_candle_count` (see the [parameter table](freqai-parameter-table.md) for detailed descriptions of these parameters) from the beginning of the desired backtesting time range.
|
||||
|
||||
As an example, to backtest the `--timerange 20210501-20210701` using the [example config](freqai-configuration.md#setting-up-the-configuration-file) which sets `train_period_days` to 30, together with `startup_candle_count: 40` on a maximum `include_timeframes` of 1h, the start date for the downloaded data needs to be `20210501` - 30 days - 40 * 1h / 24 hours = 20210330 (31.7 days earlier than the start of the desired training time range).
|
||||
|
||||
### Deciding the size of the sliding training window and backtesting duration
|
||||
|
||||
The backtesting time range is defined with the typical `--timerange` parameter in the configuration file. The duration of the sliding training window is set by `train_period_days`, whilst `backtest_period_days` is the sliding backtesting window, both in number of days (`backtest_period_days` can be
|
||||
a float to indicate sub-daily retraining in live/dry mode). In the presented [example config](freqai-configuration.md#setting-up-the-configuration-file) (found in `config_examples/config_freqai.example.json`), the user is asking FreqAI to use a training period of 30 days and backtest on the subsequent 7 days. After the training of the model, FreqAI will backtest the subsequent 7 days. The "sliding window" then moves one week forward (emulating FreqAI retraining once per week in live mode) and the new model uses the previous 30 days (including the 7 days used for backtesting by the previous model) to train. This is repeated until the end of `--timerange`. This means that if you set `--timerange 20210501-20210701`, FreqAI will have trained 8 separate models at the end of `--timerange` (because the full range comprises 8 weeks).
|
||||
|
||||
!!! Note
|
||||
Although fractional `backtest_period_days` is allowed, you should be aware that the `--timerange` is divided by this value to determine the number of models that FreqAI will need to train in order to backtest the full range. For example, by setting a `--timerange` of 10 days, and a `backtest_period_days` of 0.1, FreqAI will need to train 100 models per pair to complete the full backtest. Because of this, a true backtest of FreqAI adaptive training would take a *very* long time. The best way to fully test a model is to run it dry and let it train constantly. In this case, backtesting would take the exact same amount of time as a dry run.
|
||||
|
||||
## Defining model expirations
|
||||
|
||||
During dry/live mode, FreqAI trains each coin pair sequentially (on separate threads/GPU from the main Freqtrade bot). This means that there is always an age discrepancy between models. If you are training on 50 pairs, and each pair requires 5 minutes to train, the oldest model will be over 4 hours old. This may be undesirable if the characteristic time scale (the trade duration target) for a strategy is less than 4 hours. You can decide to only make trade entries if the model is less than a certain number of hours old by setting the `expiration_hours` in the config file:
|
||||
|
||||
```json
|
||||
"freqai": {
|
||||
"expiration_hours": 0.5,
|
||||
}
|
||||
```
|
||||
|
||||
In the presented example config, the user will only allow predictions on models that are less than 1/2 hours old.
|
||||
|
||||
## Controlling the model learning process
|
||||
|
||||
Model training parameters are unique to the selected machine learning library. FreqAI allows you to set any parameter for any library using the `model_training_parameters` dictionary in the config. The example config (found in `config_examples/config_freqai.example.json`) shows some of the example parameters associated with `Catboost` and `LightGBM`, but you can add any parameters available in those libraries or any other machine learning library you choose to implement.
|
||||
|
||||
Data split parameters are defined in `data_split_parameters` which can be any parameters associated with scikit-learn's `train_test_split()` function. `train_test_split()` has a parameters called `shuffle` which allows to shuffle the data or keep it unshuffled. This is particularly useful to avoid biasing training with temporally auto-correlated data. More details about these parameters can be found the [scikit-learn website](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html) (external website).
|
||||
|
||||
The FreqAI specific parameter `label_period_candles` defines the offset (number of candles into the future) used for the `labels`. In the presented [example config](freqai-configuration.md#setting-up-the-configuration-file), the user is asking for `labels` that are 24 candles in the future.
|
||||
|
||||
## Continual learning
|
||||
|
||||
You can choose to adopt a continual learning scheme by setting `"continual_learning": true` in the config. By enabling `continual_learning`, after training an initial model from scratch, subsequent trainings will start from the final model state of the preceding training. This gives the new model a "memory" of the previous state. By default, this is set to `False` which means that all new models are trained from scratch, without input from previous models.
|
||||
|
||||
???+ danger "Continual learning enforces a constant parameter space"
|
||||
Since `continual_learning` means that the model parameter space *cannot* change between trainings, `principal_component_analysis` is automatically disabled when `continual_learning` is enabled. Hint: PCA changes the parameter space and the number of features, learn more about PCA [here](freqai-feature-engineering.md#data-dimensionality-reduction-with-principal-component-analysis).
|
||||
|
||||
???+ danger "Experimental functionality"
|
||||
Beware that this is currently a naive approach to incremental learning, and it has a high probability of overfitting/getting stuck in local minima while the market moves away from your model. We have the mechanics available in FreqAI primarily for experimental purposes and so that it is ready for more mature approaches to continual learning in chaotic systems like the crypto market.
|
||||
|
||||
## Hyperopt
|
||||
|
||||
You can hyperopt using the same command as for [typical Freqtrade hyperopt](hyperopt.md):
|
||||
|
||||
```bash
|
||||
freqtrade hyperopt --hyperopt-loss SharpeHyperOptLoss --strategy FreqaiExampleStrategy --freqaimodel LightGBMRegressor --strategy-path freqtrade/templates --config config_examples/config_freqai.example.json --timerange 20220428-20220507
|
||||
```
|
||||
|
||||
`hyperopt` requires you to have the data pre-downloaded in the same fashion as if you were doing [backtesting](#backtesting). In addition, you must consider some restrictions when trying to hyperopt FreqAI strategies:
|
||||
|
||||
- The `--analyze-per-epoch` hyperopt parameter is not compatible with FreqAI.
|
||||
- It's not possible to hyperopt indicators in the `feature_engineering_*()` and `set_freqai_targets()` functions. This means that you cannot optimize model parameters using hyperopt. Apart from this exception, it is possible to optimize all other [spaces](hyperopt.md#running-hyperopt-with-smaller-search-space).
|
||||
- The backtesting instructions also apply to hyperopt.
|
||||
|
||||
The best method for combining hyperopt and FreqAI is to focus on hyperopting entry/exit thresholds/criteria. You need to focus on hyperopting parameters that are not used in your features. For example, you should not try to hyperopt rolling window lengths in the feature creation, or any part of the FreqAI config which changes predictions. In order to efficiently hyperopt the FreqAI strategy, FreqAI stores predictions as dataframes and reuses them. Hence the requirement to hyperopt entry/exit thresholds/criteria only.
|
||||
|
||||
A good example of a hyperoptable parameter in FreqAI is a threshold for the [Dissimilarity Index (DI)](freqai-feature-engineering.md#identifying-outliers-with-the-dissimilarity-index-di) `DI_values` beyond which we consider data points as outliers:
|
||||
|
||||
```python
|
||||
di_max = IntParameter(low=1, high=20, default=10, space='buy', optimize=True, load=True)
|
||||
dataframe['outlier'] = np.where(dataframe['DI_values'] > self.di_max.value/10, 1, 0)
|
||||
```
|
||||
|
||||
This specific hyperopt would help you understand the appropriate `DI_values` for your particular parameter space.
|
||||
|
||||
## Using Tensorboard
|
||||
|
||||
!!! note "Availability"
|
||||
FreqAI includes tensorboard for a variety of models, including XGBoost, all PyTorch models, Reinforcement Learning, and Catboost. If you would like to see Tensorboard integrated into another model type, please open an issue on the [Freqtrade GitHub](https://github.com/freqtrade/freqtrade/issues)
|
||||
|
||||
!!! danger "Requirements"
|
||||
Tensorboard logging requires the FreqAI torch installation/docker image.
|
||||
|
||||
|
||||
The easiest way to use tensorboard is to ensure `freqai.activate_tensorboard` is set to `True` (default setting) in your configuration file, run FreqAI, then open a separate shell and run:
|
||||
|
||||
```bash
|
||||
cd freqtrade
|
||||
tensorboard --logdir user_data/models/unique-id
|
||||
```
|
||||
|
||||
where `unique-id` is the `identifier` set in the `freqai` configuration file. This command must be run in a separate shell if you wish to view the output in your browser at 127.0.0.1:6060 (6060 is the default port used by Tensorboard).
|
||||
|
||||
|
||||
|
||||
|
||||
!!! note "Deactivate for improved performance"
|
||||
Tensorboard logging can slow down training and should be deactivated for production use.
|
||||
141
md/freqai.md
Normal file
141
md/freqai.md
Normal file
@ -0,0 +1,141 @@
|
||||
|
||||
|
||||
# FreqAI
|
||||
|
||||
## Introduction
|
||||
|
||||
FreqAI is a software designed to automate a variety of tasks associated with training a predictive machine learning model to generate market forecasts given a set of input signals. In general, FreqAI aims to be a sandbox for easily deploying robust machine learning libraries on real-time data ([details](#freqai-position-in-open-source-machine-learning-landscape)).
|
||||
|
||||
!!! Note
|
||||
FreqAI is, and always will be, a not-for-profit, open-source project. FreqAI does *not* have a crypto token, FreqAI does *not* sell signals, and FreqAI does not have a domain besides the present [freqtrade documentation](https://www.freqtrade.io/en/latest/freqai/).
|
||||
|
||||
Features include:
|
||||
|
||||
* **Self-adaptive retraining** - Retrain models during [live deployments](freqai-running.md#live-deployments) to self-adapt to the market in a supervised manner
|
||||
* **Rapid feature engineering** - Create large rich [feature sets](freqai-feature-engineering.md#feature-engineering) (10k+ features) based on simple user-created strategies
|
||||
* **High performance** - Threading allows for adaptive model retraining on a separate thread (or on GPU if available) from model inferencing (prediction) and bot trade operations. Newest models and data are kept in RAM for rapid inferencing
|
||||
* **Realistic backtesting** - Emulate self-adaptive training on historic data with a [backtesting module](freqai-running.md#backtesting) that automates retraining
|
||||
* **Extensibility** - The generalized and robust architecture allows for incorporating any [machine learning library/method](freqai-configuration.md#using-different-prediction-models) available in Python. Eight examples are currently available, including classifiers, regressors, and a convolutional neural network
|
||||
* **Smart outlier removal** - Remove outliers from training and prediction data sets using a variety of [outlier detection techniques](freqai-feature-engineering.md#outlier-detection)
|
||||
* **Crash resilience** - Store trained models to disk to make reloading from a crash fast and easy, and [purge obsolete files](freqai-running.md#purging-old-model-data) for sustained dry/live runs
|
||||
* **Automatic data normalization** - [Normalize the data](freqai-feature-engineering.md#building-the-data-pipeline) in a smart and statistically safe way
|
||||
* **Automatic data download** - Compute timeranges for data downloads and update historic data (in live deployments)
|
||||
* **Cleaning of incoming data** - Handle NaNs safely before training and model inferencing
|
||||
* **Dimensionality reduction** - Reduce the size of the training data via [Principal Component Analysis](freqai-feature-engineering.md#data-dimensionality-reduction-with-principal-component-analysis)
|
||||
* **Deploying bot fleets** - Set one bot to train models while a fleet of [consumers](producer-consumer.md) use signals.
|
||||
|
||||
## Quick start
|
||||
|
||||
The easiest way to quickly test FreqAI is to run it in dry mode with the following command:
|
||||
|
||||
```bash
|
||||
freqtrade trade --config config_examples/config_freqai.example.json --strategy FreqaiExampleStrategy --freqaimodel LightGBMRegressor --strategy-path freqtrade/templates
|
||||
```
|
||||
|
||||
You will see the boot-up process of automatic data downloading, followed by simultaneous training and trading.
|
||||
|
||||
!!! danger "Not for production"
|
||||
The example strategy provided with the Freqtrade source code is designed for showcasing/testing a wide variety of FreqAI features. It is also designed to run on small computers so that it can be used as a benchmark between developers and users. It is *not* designed to be run in production.
|
||||
|
||||
An example strategy, prediction model, and config to use as a starting points can be found in
|
||||
`freqtrade/templates/FreqaiExampleStrategy.py`, `freqtrade/freqai/prediction_models/LightGBMRegressor.py`, and
|
||||
`config_examples/config_freqai.example.json`, respectively.
|
||||
|
||||
## General approach
|
||||
|
||||
You provide FreqAI with a set of custom *base indicators* (the same way as in a [typical Freqtrade strategy](strategy-customization.md)) as well as target values (*labels*). For each pair in the whitelist, FreqAI trains a model to predict the target values based on the input of custom indicators. The models are then consistently retrained, with a predetermined frequency, to adapt to market conditions. FreqAI offers the ability to both backtest strategies (emulating reality with periodic retraining on historic data) and deploy dry/live runs. In dry/live conditions, FreqAI can be set to constant retraining in a background thread to keep models as up to date as possible.
|
||||
|
||||
An overview of the algorithm, explaining the data processing pipeline and model usage, is shown below.
|
||||
|
||||
|
||||
|
||||
### Important machine learning vocabulary
|
||||
|
||||
**Features** - the parameters, based on historic data, on which a model is trained. All features for a single candle are stored as a vector. In FreqAI, you build a feature data set from anything you can construct in the strategy.
|
||||
|
||||
**Labels** - the target values that the model is trained toward. Each feature vector is associated with a single label that is defined by you within the strategy. These labels intentionally look into the future and are what you are training the model to be able to predict.
|
||||
|
||||
**Training** - the process of "teaching" the model to match the feature sets to the associated labels. Different types of models "learn" in different ways which means that one might be better than another for a specific application. More information about the different models that are already implemented in FreqAI can be found [here](freqai-configuration.md#using-different-prediction-models).
|
||||
|
||||
**Train data** - a subset of the feature data set that is fed to the model during training to "teach" the model how to predict the targets. This data directly influences weight connections in the model.
|
||||
|
||||
**Test data** - a subset of the feature data set that is used to evaluate the performance of the model after training. This data does not influence nodal weights within the model.
|
||||
|
||||
**Inferencing** - the process of feeding a trained model new unseen data on which it will make a prediction.
|
||||
|
||||
## Install prerequisites
|
||||
|
||||
The normal Freqtrade install process will ask if you wish to install FreqAI dependencies. You should reply "yes" to this question if you wish to use FreqAI. If you did not reply yes, you can manually install these dependencies after the install with:
|
||||
|
||||
``` bash
|
||||
pip install -r requirements-freqai.txt
|
||||
```
|
||||
|
||||
!!! Note
|
||||
Catboost will not be installed on low-powered arm devices (raspberry), since it does not provide wheels for this platform.
|
||||
|
||||
### Usage with docker
|
||||
|
||||
If you are using docker, a dedicated tag with FreqAI dependencies is available as `:freqai`. As such - you can replace the image line in your docker compose file with `image: freqtradeorg/freqtrade:stable_freqai`. This image contains the regular FreqAI dependencies. Similar to native installs, Catboost will not be available on ARM based devices. If you would like to use PyTorch or Reinforcement learning, you should use the torch or RL tags, `image: freqtradeorg/freqtrade:stable_freqaitorch`, `image: freqtradeorg/freqtrade:stable_freqairl`.
|
||||
|
||||
!!! note "docker-compose-freqai.yml"
|
||||
We do provide an explicit docker-compose file for this in `docker/docker-compose-freqai.yml` - which can be used via `docker compose -f docker/docker-compose-freqai.yml run ...` - or can be copied to replace the original docker file. This docker-compose file also contains a (disabled) section to enable GPU resources within docker containers. This obviously assumes the system has GPU resources available.
|
||||
|
||||
### FreqAI position in open-source machine learning landscape
|
||||
|
||||
Forecasting chaotic time-series based systems, such as equity/cryptocurrency markets, requires a broad set of tools geared toward testing a wide range of hypotheses. Fortunately, a recent maturation of robust machine learning libraries (e.g. `scikit-learn`) has opened up a wide range of research possibilities. Scientists from a diverse range of fields can now easily prototype their studies on an abundance of established machine learning algorithms. Similarly, these user-friendly libraries enable "citizen scientists" to use their basic Python skills for data exploration. However, leveraging these machine learning libraries on historical and live chaotic data sources can be logistically difficult and expensive. Additionally, robust data collection, storage, and handling presents a disparate challenge. [`FreqAI`](#freqai) aims to provide a generalized and extensible open-sourced framework geared toward live deployments of adaptive modeling for market forecasting. The `FreqAI` framework is effectively a sandbox for the rich world of open-source machine learning libraries. Inside the `FreqAI` sandbox, users find they can combine a wide variety of third-party libraries to test creative hypotheses on a free live 24/7 chaotic data source - cryptocurrency exchange data.
|
||||
|
||||
### Citing FreqAI
|
||||
|
||||
FreqAI is [published in the Journal of Open Source Software](https://joss.theoj.org/papers/10.21105/joss.04864). If you find FreqAI useful in your research, please use the following citation:
|
||||
|
||||
```bibtex
|
||||
@article{Caulk2022,
|
||||
doi = {10.21105/joss.04864},
|
||||
url = {https://doi.org/10.21105/joss.04864},
|
||||
year = {2022}, publisher = {The Open Journal},
|
||||
volume = {7}, number = {80}, pages = {4864},
|
||||
author = {Robert A. Caulk and Elin Törnquist and Matthias Voppichler and Andrew R. Lawless and Ryan McMullan and Wagner Costa Santos and Timothy C. Pogue and Johan van der Vlugt and Stefan P. Gehring and Pascal Schmidt},
|
||||
title = {FreqAI: generalizing adaptive modeling for chaotic time-series market forecasts},
|
||||
journal = {Journal of Open Source Software} }
|
||||
```
|
||||
|
||||
## Common pitfalls
|
||||
|
||||
FreqAI cannot be combined with dynamic `VolumePairlists` (or any pairlist filter that adds and removes pairs dynamically).
|
||||
This is for performance reasons - FreqAI relies on making quick predictions/retrains. To do this effectively,
|
||||
it needs to download all the training data at the beginning of a dry/live instance. FreqAI stores and appends
|
||||
new candles automatically for future retrains. This means that if new pairs arrive later in the dry run due to a volume pairlist, it will not have the data ready. However, FreqAI does work with the `ShufflePairlist` or a `VolumePairlist` which keeps the total pairlist constant (but reorders the pairs according to volume).
|
||||
|
||||
## Additional learning materials
|
||||
|
||||
Here we compile some external materials that provide deeper looks into various components of FreqAI:
|
||||
|
||||
- [Real-time head-to-head: Adaptive modeling of financial market data using XGBoost and CatBoost](https://emergentmethods.medium.com/real-time-head-to-head-adaptive-modeling-of-financial-market-data-using-xgboost-and-catboost-995a115a7495)
|
||||
- [FreqAI - from price to prediction](https://emergentmethods.medium.com/freqai-from-price-to-prediction-6fadac18b665)
|
||||
|
||||
|
||||
## Support
|
||||
|
||||
You can find support for FreqAI in a variety of places, including the [Freqtrade discord](https://discord.gg/Jd8JYeWHc4), the dedicated [FreqAI discord](https://discord.gg/7AMWACmbjT), and in [github issues](https://github.com/freqtrade/freqtrade/issues).
|
||||
|
||||
## Credits
|
||||
|
||||
FreqAI is developed by a group of individuals who all contribute specific skillsets to the project.
|
||||
|
||||
Conception and software development:
|
||||
Robert Caulk @robcaulk
|
||||
|
||||
Theoretical brainstorming and data analysis:
|
||||
Elin Törnquist @th0rntwig
|
||||
|
||||
Code review and software architecture brainstorming:
|
||||
@xmatthias
|
||||
|
||||
Software development:
|
||||
Wagner Costa @wagnercosta
|
||||
Emre Suzen @aemr3
|
||||
Timothy Pogue @wizrds
|
||||
|
||||
Beta testing and bug reporting:
|
||||
Stefan Gehring @bloodhunter4rc, @longyu, Andrew Lawless @paranoidandy, Pascal Schmidt @smidelis, Ryan McMullan @smarmau, Juha Nykänen @suikula, Johan van der Vlugt @jooopiert, Richárd Józsa @richardjosza
|
||||
Loading…
x
Reference in New Issue
Block a user