李辉, 化金金, 邹波蓉.基于RF-LSTM组合模型的股票价格预测[J].河南理工大学学报（自然科学版）,2022,41(1):136-142.

LI H, HUA J J, ZOU B R. Stock price prediction based on RF-LSTM combination model[J].Journal of Henan Polytechnic University(Natural Science) ,2022,41(1):136-142.

基于RF-LSTM组合模型的股票价格预测

李辉1, 化金金2, 邹波蓉1

1.河南理工大学 物理与电子信息学院,河南 焦作454000；2.河南理工大学 电气工程与自动化学院,河南 焦作 454000

摘要:股票数据具有非线性和复杂性等特点,单一模型预测效果不佳,针对此问题,提出一种 RF-LSTM组合模型,用于预测股票的收盘价。首先,利用Tushare财经数据包获取股票数据,构建特征集,并对数据进行归一化处理；其次,考虑到多特征之间存在高度的非线性和信息冗余问题,利用随机森林（RF）选择最优特征集,降低数据维度和训练复杂度；最后,利用深度学习中适合处理时间序列的长短期记忆网络（LSTM）对股票价格进行预测,并对预测模型进行参数调优。结果表明,与单一结构的LSTM神经网络模型预测相比,本文提出的RF-LSTM组合模型预测的平均绝对误差（M4E ）、均方误差（MSE）和均方根误差（RMSE）分别减小了 13. 11％,6.70％和12.54％。该组合模型可提高股票价格预测的准确性。

关键词:股票预测;机器学习;长短时记忆网络;随机森林

doi:10.16186/j.cnki.1673-9787.2019100021

基金项目:河南省基础与前沿技术研究计划项目（152300410103）

收稿日期:2020/08/10

修回日期:2020/12/02

出版日期:2022/01/01

Stock price prediction based on RF-LSTM combination model

LI Hui1, HUA Jinjin2, ZOU Borong1

1.School of Physics and Electronic Information Engineering, Henan Polytechnic University, Jiaozuo 454000 , Henan, China；2.School of Electrical Engineering and Automation, Henan Polytechnic University, Jiaozuo  454000 , Henan, China

Abstract: Stock data is characterized by non-linear complexity,and a single model has poor prediction results.Aiming to the above problems,RF-LSTM combination model was proposed to predict the closing price of stocks.First,the Tushare financial data package was used to obtain stock data,the feature sets were built and the data was normalized.Secondly,considering the high non-linearity and information redundancy between multiple features,the random forest(RF) was used to select the optimal feature sets and to reduce the dimension of data and the complexity of training.Then,the long-short term memory(LSTM) ,which was suitable for processing time series in deep learning,was used to predict stock prices and to optimize the parameters of the prediction model.The experimental results showed that compared with the single structure LSTM neural network model,the mean absolute error(MAE) ,mean square error(MSE) and root mean square error(RMSE) of the combined RF-LSTM model were reduced by 13.11% ,6.70% ,and 12.54% respectively.The combination model could increase the predicton accuracy of stockprice.

Key words:stockprice prediction;machine learning;long-short term memory;random forest

 基于RF-LSTM组合模型的股票价格预测_李辉.pdf