CRYPTO CURRENCY ALPHA GENERATOR USING MACHINE LEARNING

Aditya Ravindra Pawar; Anish Ajay Zuting; Hardik Jain; Arjun Dhananjay Shedage; Ashvini Shirke

Authors

Aditya Ravindra Pawar STUDENT , Dr D.Y. Patil College of Engineering & Innovation
Author
Anish Ajay Zuting STUDENT, Dr D.Y. Patil College of Engineering & Innovation
Author
Hardik Jain STUDENT , Dr D.Y. Patil College of Engineering & Innovation
Author
Arjun Dhananjay Shedage STUDENT, Dr D.Y. Patil College of Engineering & Innovation
Author
Ashvini Shirke GUIDE , Dr D.Y. Patil College of Engineering & Innovation
Author

DOI:

Keywords:

Cryptocurrency Trading, Machine Learning, Logistic Regression, Mean Reversion, Trend Following, K-Means Clustering, Random Forest, Volatility Regime Detection, Algorithmic Trading, Walk-Forward Validation.

Abstract

This research presents a machine-learning-based Cryptocurrency Alpha Generator integrating Z-score mean reversion and Donchian-channel trend-following strategies with Logistic Regression meta-filtering. K-Means clustering identifies volatility regimes, while Random Forest regression supports volatility-based position sizing. The framework is evaluated across five cryptocurrency pairs and six timeframes using historical Binance market data and a 0.30% modeled round-trip transaction cost. Results demonstrate improved risk-adjusted performance in selected configurations, with SOL/USDT 15-minute mean reversion achieving a Sharpe ratio of 1.00 and a 69.17% win rate.