Predicting Monthly Returns in A-Share Market Using Supervised Learning
Description: This GitHub repository focuses on applying classical supervised learning techniques to tackle the challenge of predicting asset returns in a high-dimensional context, particularly in the A-share stock market. Considering the unique characteristics of financial data such as low signal-to-noise ratio and high-dimensional forecasting, this project employs regularization methods for feature selection. With carefully designed loss functions and cross-validation strategies, it addresses the prediction of portfolio performance as well as the issue of information leakage across different time periods.
Through a series of iterations, starting from Ordinary Least Squares (OLS), Ridge Regression, and Lasso models, this repository aims to enhance predictive performance for monthly return forecasts in the A-share market. The analysis covers the period from January 2011 to March 2023, and it systematically introduces and refines various modeling approaches. Both in-sample and out-of-sample evaluations are conducted to observe improvements in predictive accuracy.
The project emphasizes reproducibility and the application of machine learning techniques to real-world financial data, serving as a valuable resource for researchers and practitioners interested in asset pricing and prediction in high-dimensional financial scenarios.