基于机器学习的股票收益预测与投资组合研究

Machine Learning Based Stock Return Prediction and Portfolio Research

摘要: 计算机和互联网的高速发展使得量化投资在全球逐渐兴起。笔者将机器学习模型和多因子模型相结合构建量化选股模型，并使用上证50指数成分股2016年到2022年的日频数据进行模型训练和样本外预测，结果发现：1) 以随机森林、支持向量机、XGBoost三个模型进行选股构建的投资策略能够战胜市场；2) 投资收益受市场行情影响巨大，在下跌行情中，主动型投资策略即使能够战胜市场，也不能保证获得超过无风险收益率的收益。

Abstract: The rapid development of computers and the Internet has led to the gradual rise of quantitative investment worldwide. This author combines machine learning models and multi-factor models to construct a quantitative stock selection model, and uses the daily frequency data of the constituents of the SSE 50 index from 2016 to 2022 for model training and out-of-sample prediction, and finds that 1) The investment strategy constructed by stock selection with the three models of Random Forests, Support Vector Machines, and XGBoost is able to the market; 2) The investment return is affected by the market sentiment greatly, and it is difficult to get more than the risk-free rate of return in the falling market.