This project focuses on forecasting Bitcoin price volatility using ensemble-based machine learning models applied to historical market data. The goal is to evaluate the effectiveness of tree-based models in capturing non-linear dynamics in highly volatile cryptocurrency markets.
Cryptocurrency markets exhibit extreme volatility and non-stationary behavior, making traditional forecasting approaches challenging. This project aims to:
- Model short-term volatility patterns
- Compare ensemble learning approaches for predictive performance
- Collected historical Bitcoin OHLCV data
- Engineered lag-based, rolling-window, and statistical features
- Trained ensemble regression models including Random Forest and XGBoost
- Performed hyperparameter tuning to reduce prediction error variance
- Evaluated models using MAE and R² metrics
- Achieved a Mean Absolute Error (MAE) of approximately 1,250 USD
- Achieved an R² score of 0.92 on held-out evaluation data
- Reduced prediction error variance by ~18% through feature optimization and tuning
- Python
- Pandas, NumPy
- Scikit-learn
- XGBoost
- Matplotlib / Seaborn
✅ Core modeling completed
🚧 Ongoing improvements:
- Model comparison across market regimes
- Deployment as an interactive analytics dashboard
This project is for educational purposes only and does not constitute financial or investment advice.