53 lines
1.6 KiB
Python
53 lines
1.6 KiB
Python
|
"""
|
||
|
|
Convert ML V3 model to TradingModelV2 compatible format.
|
||
|
|
"""
|
||
|
|
|
||
|
|
import pickle
|
||
|
|
import sys
|
||
|
|
from pathlib import Path
|
||
|
|
|
||
|
|
# Add project root to path
|
||
|
|
sys.path.insert(0, str(Path(__file__).parent.parent.parent))
|
||
|
|
|
||
|
|
from backtests.ml_v2.ml_v2_model import ModelType
|
||
|
|
|
||
|
|
# Load old format
|
||
|
|
old_path = Path("backtests/ml_v3/xgboost_model_v3.pkl")
|
||
|
|
with open(old_path, 'rb') as f:
|
||
|
|
old_data = pickle.load(f)
|
||
|
|
|
||
|
|
print(f"Loaded model from: {old_path}")
|
||
|
|
print(f"Old keys: {list(old_data.keys())}")
|
||
|
|
|
||
|
|
# Convert to TradingModelV2 format
|
||
|
|
new_data = {
|
||
|
|
'xgb_model': old_data['model'], # XGBoost Booster object
|
||
|
|
'lgb_model': None,
|
||
|
|
'model_type': ModelType.XGBOOST_BINARY,
|
||
|
|
'feature_names': old_data['feature_cols'],
|
||
|
|
'confidence_threshold': 0.60,
|
||
|
|
'xgb_params': old_data['metadata'].get('hyperparameters', {}),
|
||
|
|
'lgb_params': {},
|
||
|
|
'feature_importance': {},
|
||
|
|
'train_metrics': {
|
||
|
|
'train_accuracy': old_data['metadata']['train_accuracy'],
|
||
|
|
'test_accuracy': old_data['metadata']['test_accuracy'],
|
||
|
|
},
|
||
|
|
'fitted': True,
|
||
|
|
'metadata': old_data['metadata'],
|
||
|
|
'version': '3.0_binary',
|
||
|
|
'trained_at': old_data['trained_at'],
|
||
|
|
'symbol': old_data['symbol'],
|
||
|
|
'timeframe': old_data['timeframe']
|
||
|
|
}
|
||
|
|
|
||
|
|
# Save new format
|
||
|
|
with open(old_path, 'wb') as f:
|
||
|
|
pickle.dump(new_data, f)
|
||
|
|
|
||
|
|
print(f"\n✅ Model converted to TradingModelV2 format!")
|
||
|
|
print(f" Model type: {new_data['model_type'].value}")
|
||
|
|
print(f" Features: {len(new_data['feature_names'])}")
|
||
|
|
print(f" Train accuracy: {new_data['train_metrics']['train_accuracy']:.4f}")
|
||
|
|
print(f" Test accuracy: {new_data['train_metrics']['test_accuracy']:.4f}")
|