摘要
为了提高股票预测的正确率,参照股票研究的指标体系,以股票的相对强弱、变动速率、能量潮、异同移动平均线以及威廉指标五个纯技术指标作为股票预测的特征。通过网格搜索对随机森林的参数进行了优化,构建基于纯技术指标的和参数优化随机森林的股票预测模型,并以平安银行、万科、深振业A、神州高铁、美丽生态2017年4月30日到2019年6月30日所有交易日作为实验室数据,实验结果与原始随机森林、决策树以及支持向量机分类模型对比,证实了参数优化后的随机森林股票预测模型在模型评价中的准确率和AUC值都高于其他模型。
In order to improve the accuracy of stock prediction,according to the index system of stock research,RSI,ROC,OBV,MACD and Williams%R are taken as the characteristics of stock prediction.The parameters of the random forest were optimized through grid search,and a stock prediction model based on pure technical indicators and parametric optimized random forest was constructed.All trading days from April 30,2017 to June 30,2019 were taken as laboratory data by Ping An Bank,Vanke,SHENZHEN ZHENYE(GROUP)CO.LTD,China High-speed Railway and Beautiful Ecology.The experimental results are compared with the original random forest,decision tree and SVM classification models,and it is confirmed that the accuracy and AUC of the optimized random forest stock prediction model are higher than other models in model evaluation.
作者
邓晶
李路
DENG Jing;LI Lu(School of Mathematics,Physics and Statistics,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《软件》
2020年第1期178-182,共5页
Software
关键词
随机森林
技术指标
参数优化
网格搜索
股价预测
Random forest
Technical indicators
Parametric optimization
Grid search
Stock price forecasting