基于LightGBM的股票收益率预测与策略实现
Stock Return Prediction and Strategy Implementation Based on LightGBM
摘要
针对传统股票预测模型精度不足、落地性弱的痛点，本文以港股腾讯控股（00700.HK）为研究对象，采用LightGBM算法构建股票收益率预测模型，设计配套的趋势跟踪+止损量化投资策略。研究以2018—2023年日线数据为训练样本，2024—2025年为样本外回测区间。实证结果表明，模型涨跌方向判断准确率达60%，策略实现36.65%的年化收益率，最大回撤仅12.21%，显著优于买入持有基准，可为个人投资者提供低门槛的港股量化投资方案。
Abstract
Aiming at the problems of insufficient accuracy and weak practicability of traditional stock prediction models, this paper takes Tencent Holdings (00700.HK) in the Hong Kong stock market as the research object, and uses the LightGBM algorithm to construct a stock return prediction model. A corresponding quantitative investment strategy combining trend tracking and stop-loss is designed. The study uses daily data from 2018 to 2023 as the training sample, and the period from 2024 to 2025 as the out-of-sample backtesting interval. The empirical results show that the model achieves an accuracy of 60% in judging the upward and downward trends. The strategy realizes an annual return rate of 36.65% with a maximum drawdown of only 12.21%, which is significantly better than the buy-and-hold benchmark. It can provide a low-threshold quantitative investment scheme for individual investors in the Hong Kong stock market.关键词
LightGBM/股票收益率预测/量化投资策略/港股市场/机器学习应用Key words
LightGBM/stock return prediction/quantitative investment strategy/Hong Kong stock market/machine learning application引用本文复制引用
朱芝萱.基于LightGBM的股票收益率预测与策略实现[EB/OL].(2026-03-12)[2026-03-14].https://chinaxiv.org/abs/202603.00065.学科分类
财政、金融
评论