ML Ensemble BTC 5-Min Scalping (LGBM+XGB+LSTM)
Ensemble machine learning model predicting Bitcoin Up/Down direction every 5 minutes on Polymarket
This bot uses a three-model ensemble (LightGBM, XGBoost, and LSTM) to predict BTC price direction over 5-minute windows. Models are retrained automatically via GitHub Actions on a scheduled basis using live market data. The bot enters positions on the YES or NO side of 5-minute BTC Up/Down markets based on the ensemble's highest-confidence prediction. Exit is automatic at market expiry (5 minutes). The repo contains 784MB of training data and model weights, suggesting extensive historical backtesting. Implementation uses Python with the Polymarket CLOB API. Risk is high due to short timeframe and binary nature of 5-minute markets; slippage and fees must be carefully managed.