Empirical Asset Pricing via Machine Learning: Evidence from China
Download as PDF
DOI: 10.23977/ICEMGD2020.015
Corresponding Author
Yue Pei
ABSTRACT
This paper analyses the relative importance of 75 individual firm characteristics on expected earnings in China’s A-share market, using random forests model in the field of machine learning. Empirically, we found that the stock turnover, industry momentum and RMB trading volume were the three most important characteristics. In addition, according to the importance of variables, the model was reconstructed with the top 23 firm characteristics after ranking, summing and dimension reduction. In comparison with the predicted results of 75 firm characteristics, we found the performance was basically unchanged, but the calculation time was saved by 68.21%. Our empirical results indicate that the random forest algorithm is not only more accurate but also more efficient in predicting the expected returns of stock portfolios based on firm characteristics. This paper contributes to the growing asset pricing study on Chinese stock market.
KEYWORDS
Random forest, firm characteristics, Chinese stock market