Value investment with machine learning

Abstract

This thesis investigates the application of machine learning techniques to value investment strategies within the Chinese A-share market, utilizing the BigQuant platform and the advanced StockRanker model. Our research focuses on integrating fundamental and technical factors to enhance predictive accuracy and investment performance. Initial models, based solely on fundamental factors, achieved an annualized return of 10\% but exhibited significant volatility. By incorporating technical indicators, we developed a combined model that improved the annualized return to 27.8\%.

Further optimization led to the creation of the Improved Fundamental and Technical Factors Model, which achieved an impressive annualized return of 50\% and a Sharpe ratio of 1.85. After fine-tuning key parameters, the final optimized model demonstrated exceptional performance, with an annualized return of 59.82\%, a Sharpe ratio of 2.13, and a win rate of 75\%. These results highlight the effectiveness of our multi-factor approach and the robustness of the StockRanker model.

The study validates the potential of machine learning-enhanced value investment strategies in emerging markets. Future research directions include incorporating alternative data sources, developing real-time adaptive algorithms, exploring advanced feature engineering techniques, and applying the models to other emerging markets.