fix: resolve ML filter dtype error and missing BTC/ETH correlation features

- Fix LightGBM predict_proba ValueError by filtering FEATURE_COLS and casting to float64
- Extract BTC/ETH correlation data from embedded parquet columns instead of missing separate files
- Disable ONNX priority in ML filter tests to use mocked LightGBM correctly
- Add NO_ML_FILTER=true to .env.example (ML adds no value with current signal thresholds)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.env.example

@@ -6,6 +6,7 @@ LEVERAGE=10
 RISK_PER_TRADE=0.02
 DISCORD_WEBHOOK_URL=
 ML_THRESHOLD=0.55
+NO_ML_FILTER=true
 MAX_SAME_DIRECTION=2
 ATR_SL_MULT=2.0
 ATR_TP_MULT=2.0

src/backtester.py

@@ -164,9 +164,11 @@ def _get_ml_proba(ml_filter: MLFilter | None, features: pd.Series) -> float | No
             X = features[FEATURE_COLS].values.astype(np.float32).reshape(1, -1)
             return float(ml_filter._onnx_session.run(None, {input_name: X})[0][0])
         else:
-            X = features.to_frame().T
+            available = [c for c in FEATURE_COLS if c in features.index]
+            X = pd.DataFrame([features[available].values.astype(np.float64)], columns=available)
             return float(ml_filter._lgbm_model.predict_proba(X)[0][1])
-    except Exception:
+    except Exception as e:
+        logger.warning(f"ML PROBA ERROR: {e}")
         return None
 
 
@@ -209,13 +211,24 @@ class Backtester:
         all_signals: dict[str, np.ndarray] = {}
         all_features: dict[str, pd.DataFrame] = {}
 
-        # BTC/ETH 상관 데이터 (있으면 로드)
-        btc_df = self._try_load_corr("BTCUSDT")
-        eth_df = self._try_load_corr("ETHUSDT")
-
         for sym in self.cfg.symbols:
             df = _load_data(sym, self.cfg.start, self.cfg.end)
             all_data[sym] = df
+
+            # BTC/ETH 상관 데이터: 임베딩된 컬럼에서 추출 (별도 파일 폴백)
+            base_cols = ["open", "high", "low", "close", "volume"]
+            btc_df = eth_df = None
+            if "close_btc" in df.columns:
+                btc_df = df[[c + "_btc" for c in base_cols]].copy()
+                btc_df.columns = base_cols
+            else:
+                btc_df = self._try_load_corr("BTCUSDT")
+            if "close_eth" in df.columns:
+                eth_df = df[[c + "_eth" for c in base_cols]].copy()
+                eth_df.columns = base_cols
+            else:
+                eth_df = self._try_load_corr("ETHUSDT")
+
             df_ind = _calc_indicators(df)
             all_indicators[sym] = df_ind
             sig_arr = _calc_signals(

src/ml_filter.py

@@ -141,7 +141,8 @@ class MLFilter:
                 X = features[FEATURE_COLS].values.astype(np.float32).reshape(1, -1)
                 proba = float(self._onnx_session.run(None, {input_name: X})[0][0])
             else:
-                X = features.to_frame().T
+                available = [c for c in FEATURE_COLS if c in features.index]
+                X = pd.DataFrame([features[available].values.astype(np.float64)], columns=available)
                 proba = float(self._lgbm_model.predict_proba(X)[0][1])
             logger.debug(
                 f"ML 필터 [{self.active_backend}] 확률: {proba:.3f} "

tests/test_ml_filter.py

@@ -32,6 +32,7 @@ def test_no_model_should_enter_returns_true(tmp_path):
 def test_should_enter_above_threshold():
     """확률 >= 0.60 이면 True"""
     f = MLFilter(threshold=0.60)
+    f._onnx_session = None  # ONNX 비활성화, LightGBM만 테스트
     mock_model = MagicMock()
     mock_model.predict_proba.return_value = np.array([[0.35, 0.65]])
     f._lgbm_model = mock_model
@@ -42,6 +43,7 @@ def test_should_enter_above_threshold():
 def test_should_enter_below_threshold():
     """확률 < 0.60 이면 False"""
     f = MLFilter(threshold=0.60)
+    f._onnx_session = None  # ONNX 비활성화, LightGBM만 테스트
     mock_model = MagicMock()
     mock_model.predict_proba.return_value = np.array([[0.55, 0.45]])
     f._lgbm_model = mock_model