Research on Multi-Factor Stock Selection Strategies in China's A-Share Market Based on Machine Learning and Deep Learning
Haoyi Liao
Renmin University of China
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
This study evaluates multi-factor stock selection strategies using LightGBM, CNN-LSTM, and Transformer on A-share market data from November 2020 to November 2025. A 99-dimensional feature system is constructed, and strict rolling-window training ensures out-of-sample predictions. Long and long-short portfolios under equal- and valueweighted schemes are backtested and evaluated via cumulative returns, risk metrics, and CH-4 factor regressions. Results show that all models possess significant predictive power, with CNN-LSTM achieving the highest Information Coefficient. Long-short strategies outperform long-only counterparts, and the CNN-LSTM long-short strategy delivers optimal riskadjusted returns, constituting an 'architecture paradox' where local-temporal modeling better captures A-share information diffusion. Significant Alphas for CNN-LSTM and Transformer confirm machine learning's ability to uncover incremental pricing information beyond traditional factors.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
计算机 / AIStock Market Forecasting Methods
Financial Markets and Investment Strategies · Financial Analysis and Corporate Governance