期刊论文

Research on stock price prediction based on Xgboost algorithm with pearson optimization

中文

信息技术

1009-2552

2018

9

84-89

10.13274/j.cnki.hdzj.2018.09.019

(16A182):湖南省教育厅科研重点项目

本校为第一机构

针对以往股票预测模型由于数据庞杂引起的低准确率以及传统神经网络训练复杂耗时的缺陷,文中提出一种结合皮尔森特征筛选与Xgboost算法的预测方法,用于预测股票收盘价。文中以A股市场600677股票为例,采用皮尔森相关系数分析法提取庞杂数据中最影响收盘价走势的相关特征,生成高相关特征的数据集,基于数据挖掘算法中表现极为优异的Xgboost框架,结合特征工程处理,取得良好的训练预测效果。对照组中将公司联动性影响因素融合原数据生成训练集,完成特征扩充,以实现与特征精简方法对比。实验结果表明在股票预测分析中,使用特征精方法简保留4项最相关属性,可有效改善预测效果,在股价走势分析中精简分析...

Aiming at the low accuracy rate and the complexity and time consuming of traditional neural network training,a prediction method combining Pearson feature selection and Xgboost algorithm is proposed to predict stock closing price. Taking the A stock market 600677 stocks as an example,it used the Pearson correlation coefficient of extracting relevant features closing price movements that most influence the complex data analysis method,the generation of high correlation feature data set based on data mining algorithm performance is very excellent Xgboost framework,combined with the characteristi...