ML Ensemble 5-Minute BTC Prediction

This strategy uses a powerful ensemble of machine learning models—LightGBM, XGBoost, and LSTM—to predict the direction of Bitcoin for 5-minute Polymarket markets. The models are combined to improve accuracy and robustness over any single model.

The bot is designed to be self-training, automatically retraining on new data via GitHub Actions. This allows the strategy to adapt to changing market dynamics without manual intervention. The live dashboard provides real-time tracking of signals and performance.

This is a sophisticated approach that leverages cutting-edge ML techniques for time-series prediction. The ensemble architecture helps to mitigate the risk of any single model misinterpreting market conditions. The repository is large, containing model weights and training datasets for the LSTM component.

The edge is derived from the predictive power of the ML models and the ability to continuously retrain them. Risks include model overfitting, data quality issues, and the challenge of translating high-level model probabilities into profitable trade entries and exits in a real-time, low-latency trading environment.