Enhancing stock return prediction in the Chinese market: A GAN-based approach

With the rapid advancement of machine learning technologies, the accuracy of predictive models has seen continuous improvement. In this paper, we aim to apply machine learning models to predict stock market returns and implement factor investment in the Chinese A-share market. We propose a novel GAN model integrating LSTM, attention mechanisms, and CNN to address time-dependence issues and low signal-to-noise ratios more effectively in stock return prediction. Out-of-sample results demonstrate that the proposed GAN model outperforms other benchmark models in terms of predictive performance and exhibits superior generalization capabilities. Furthermore, our analysis of predictor importance highlights the pivotal role of transaction and profitability metrics in predicting returns in the Chinese A-share market. Additionally, when incorporating macroeconomic predictors into the training data, our model demonstrates stability amidst macroeconomic fluctuations. These empirical findings underscore the potential value of the proposed GAN model in the Chinese stock market investment.