Research on Importance Fluctuation of Stock Selection Factors Based on the Perspective of COVID-19 Epidemic Normalization

Abstract

Since the outbreak of COVID-19 epidemic, the efficiency of the Chinese A-share market has been severely decreased, so it is worthwhile for investors to set up a new stock selection system. This article proposes a novel factor quantitative method based on information entropy and Critic-weight for the Chinese A-share market before and after the normalization of epidemic. Then we use the combination machine learning method including Stochastic Forest, XGBoost, LightGBM, and mutual information to fit the stock return and evaluate factor importance. Under the condition that the model has stability and effectiveness, the results by using paired T-test method show that the impact of volatility factors on stock return is significantly enhanced, and the importance of factors such as scale and leverage is significantly descended.