我国A股股票收益率数据的聚类分析及投资策略研究

Clustering Analysis and Investment Strategy Research of A-Share Returns in China

摘要: 机器学习在股票领域的应用日益成为研究和实践的热点。通过分析相关上市公司的股票价格、财务数据、市场情绪和宏观经济因素等多维数据，机器学习算法能够建立预测模型，帮助投资者做出更明智的投资决策。对反映股价信息的个股、大盘、财务数据3类构建了共15项指标，然后采用机器学习中的K-Means聚类算法对我国A股收益数据进行了聚类分析，分析识别出对提高股票投资获胜概率的关键性指标以及相应合适的取值范围。在得到相关结论后，先使用板块以及指数数据对结论进行了检验，得到0.8及以上的盈利概率；然后采用2015~2022历年来的历史交易数据进行了二次验证，策略累计回报率显著优于沪深300基准指数。

Abstract: Machine learning’s application in the stock market field is increasingly becoming a focal point in both research and practice. By analyzing multidimensional data such as stock prices, financial data, market sentiments, and macroeconomic factors related to listed companies, machine learning algorithms can establish predictive models to assist investors in making wiser investment decisions. This article constructs 15 indicators across three categories—individual stocks, market indices, and financial data—to reflect stock price information. Then, using the K-Means clustering algorithm in machine learning, it conducts cluster analysis on the returns data of A-shares in China, identifying crucial indicators and their appropriate value ranges that enhance the probability of successful stock investments. After obtaining these conclusions, the study validates them initially using sector and index data, achieving a profitability probability of 0.8 or higher. Subsequently, historical trading data from 2015 to 2022 is used for a secondary validation, showing significantly better cumulative returns compared to the benchmark Shanghai and Shenzhen 300 Index.