with v3

config_examples/config_freqai.okx.json

@@ -77,7 +77,10 @@
         "live_retrain_hours": 0,
         "feature_selection": {
             "method": "recursive_elimination",
-            "threshold": 0.01
+            "threshold": 0.01,
+            "steps": 1,
+            "cv": 5,
+            "n_features_to_select": null
         },
         "feature_parameters": {
             "include_timeframes": ["3m", "5m", "1h"],
@@ -93,8 +96,8 @@
             "random_state": 42
         },
         "model_training_parameters": {
-            "n_estimators": 200,
-            "learning_rate": 0.05,
+            "n_estimators": 100,
+            "learning_rate": 0.1,
             "max_depth": 5,
             "subsample": 0.8,
             "colsample_bytree": 0.8,

freqtrade/templates/FreqaiExampleStrategy.py

@@ -72,18 +72,27 @@ class FreqaiExampleStrategy(IStrategy):
     }
 
     def featcaure_engineering_expand_all(self, dataframe: DataFrame, period: int, metadata: dict, **kwargs) -> DataFrame:
-        # 保留关键的技术指标
-        dataframe["rsi"] = ta.RSI(dataframe, timeperiod=14)
+        # 只计算必要的技术指标
+        if len(dataframe) > 14:
+            dataframe["rsi"] = ta.RSI(dataframe, timeperiod=14)
+            dataframe["ema_12"] = ta.EMA(dataframe, timeperiod=12)
+            dataframe["ema_26"] = ta.EMA(dataframe, timeperiod=26)
+        else:
+            dataframe["rsi"] = 50
+            dataframe["ema_12"] = dataframe["close"]
+            dataframe["ema_26"] = dataframe["close"]
 
         # 确保 MACD 列被正确计算并保留
-        try:
+        # 确保有足够的数据计算 MACD
+        if len(dataframe) >= 50:
             macd = ta.MACD(dataframe, fastperiod=12, slowperiod=26, signalperiod=9)
             dataframe["macd"] = macd["macd"]
             dataframe["macdsignal"] = macd["macdsignal"]
-        except Exception as e:
-            logger.error(f"计算 MACD 列时出错：{str(e)}")
-            dataframe["macd"] = np.nan
-            dataframe["macdsignal"] = np.nan
+            dataframe["macdhist"] = macd["macdhist"]
+        else:
+            dataframe["macd"] = 0
+            dataframe["macdsignal"] = 0
+            dataframe["macdhist"] = 0
 
         # 检查 MACD 列是否存在
         if "macd" not in dataframe.columns or "macdsignal" not in dataframe.columns:
@@ -148,9 +157,12 @@ class FreqaiExampleStrategy(IStrategy):
             label_period = self.freqai_info["feature_parameters"]["label_period_candles"]
 
             # 定义目标变量为未来价格变化百分比（连续值）
-            dataframe["up_or_down"] = (
-                dataframe["close"].shift(-label_period) - dataframe["close"]
-            ) / dataframe["close"]
+            # 使用对数收益率作为目标变量
+            dataframe["target"] = np.log(
+                dataframe["close"].shift(-label_period) / dataframe["close"]
+            )
+            # 处理异常值
+            dataframe["target"] = dataframe["target"].clip(-0.1, 0.1)
 
             # 数据清理：处理 NaN 和 Inf 值
             dataframe["up_or_down"] = dataframe["up_or_down"].replace([np.inf, -np.inf], np.nan)
@@ -225,9 +237,9 @@ class FreqaiExampleStrategy(IStrategy):
             
             # 简化动态参数生成逻辑
         # 放松 buy_rsi 和 sell_rsi 的生成逻辑
-            # 计算 buy_rsi_pred 并清理 NaN 值
-            dataframe["buy_rsi_pred"] = dataframe["rsi"].rolling(window=10).mean().clip(30, 50)
-            dataframe["buy_rsi_pred"] = dataframe["buy_rsi_pred"].fillna(dataframe["buy_rsi_pred"].median())
+            # 使用固定 RSI 阈值
+            self.buy_rsi.value = 30
+            self.sell_rsi.value = 70
 
             # 计算 sell_rsi_pred 并清理 NaN 值
             dataframe["sell_rsi_pred"] = dataframe["buy_rsi_pred"] + 20