Year: 2024
Type: conference-paper
Abstract: This paper studies the application of XGBoost model in the stock market. The main content is to analyze its prediction of stock returns and measure the importance of features. This paper uses the A-share data from 2001 to 2022 and constructs an XGBoost model with multiple prediction features. It is found that this model is significantly better than the traditional OLS model in predicting the excess return of stocks. For the test period from 2014 to 2022, the XGBoost strategy can achieve 155% higher returns and risk-adjusted performance than the OLS strategy in the portfolio based on the model forecast, and XGBoost has higher average monthly returns in both long and long-short strategies, and has more advantages in risk control. In particular, the maximum retracement in long-short strategies is low. Based on the feature importance analysis, it is concluded that corporate profitability indicators and medium-term momentum play the most important role in predicting the return characteristics of Chinese stocks, and technical indicators and market microstructure factors also show relatively high predictive information, potentially reflecting the unique attributes of China's A-share stock market. The research results of this paper can understand the pricing mechanism of China's stock market to a certain extent, and play an important practical role in guiding the design of quantitative investment strategies and asset management practices. The research results show the sensitivity of nonlinear models to complex market dynamics and point out broad market prospects for further in-depth research on the application potential of machine learning in financial markets.
Author: Junming Gao
Institution: Carnegie Mellon University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: bronze
OpenAlex ID: W4410250655