数据挖掘技术在量化交易指标选取中的应用

Research on Selecting Quantitative Trading Indicators Based on Data Mining

摘要: 股票市场的技术分析法是近年来量化投资研究的新热点，数据挖掘作为一门交叉学科，能够大幅度提高经济学中预测类研究的效果。本文系统性地运用数据挖掘技术，挖掘影响股票涨跌的关键指标，提升量化投资中的股票收益预测模块。本文设置特定观察期与交易策略，基于2021年1月至2022年1月A股市场的17项因子，采用数据挖掘中关联规则FP-Growth算法，使用Python编程语言对上市公司股票数据进行关联分析。在支持度阈值为30%、置信度阈值为80%条件下，挖掘出35条关联规则。实证结果显示，K线形态为阳线与短长下影、每股收益高、市盈率低、市净率低、市销率低、营业总收入同比增长率高这六大特征对股票未来收益率的影响较大。同时验证了投资策路的有效性和可行性，为股票分析以及股票投资中的特征指标的关注提供有意义的参考依据。

Abstract: The technical analysis method of the stock market has become a new hot topic in quantitative investment research in recent years. As an interdisciplinary field, data mining can greatly improve the effectiveness of predictive research in economics. This article systematically applies data mining techniques to explore key indicators that affect stock price fluctuations and improve the stock return prediction module in quantitative investment. This article sets specific observation periods and trading strategies, based on 17 factors in the A-share market from January 2021 to January 2022. The FP-Growth algorithm of association rules in data mining is used to conduct association analysis on the stock data of listed companies using Python programming language. Under the conditions of a support threshold of 30% and a confidence threshold of 80%, 35 association rules were mined. The empirical results show that the K-line pattern is characterized by a positive line and a short shadow, high earnings per share, low P/E ratio, low P/B ratio, low P/S ratio, and high year-on-year growth rate of total operating revenue, which have a significant impact on the future return of stocks. At the same time, the effectiveness and feasibility of investment strategies were verified, providing meaningful reference for stock analysis and the attention to characteristic indicators in stock investment.