基于 PCA-BP 神经网络的股票价格预测研究

摘要

在股票决策问题的研究中,针对影响股票价格因素间存在高度的非线性、存在数据冗余等特征,传统股票预测方法无法消除数据之间冗余和捕捉非线性规律导致预测精度较低,为了提高股票价格预测精度,提出一个基于主成份分析(PCA)的 BP 神经网络(BPNN)股票预测模型(PCA-BPNN).首先对影响股票价格波动的各因素进行主成份分析,消除各因素之间的冗余性,降低 BP 神经网络的输入维数,加快 BP 神经网络测速度并提高预测精度,然后利用 BPNN 对保留成分进行建模预测.利用 PEA-BPNN 模型对上海证券交易所上市的首创股份(600008)经济数据进行了验证性测试和分析,结果表明,PEA-BPNN 模型预测精度显著提高,是一种高效和准确的股票预测模型.%In stocks decision-making study, there are lots of stock prices factors and there are highly nonlinear and redundant information among factors, the traditional stock prediction method cannot eliminate data redundancy and capture nonlinear feature, and prediction accuracy is lower. In order to improve the prediction accuracy of stock prices, a stock prediction model (PCA-BPNN) is put forward based on principal component analysis (PCA) and BP neural network (BPNN). Firstly, PCA is used for selecting variables to reduce the complexity of BPNN model, eliminating redundancy among the factors, reducing the input data, and improving the speed of BPNN. By using BPNN,the model was built with the reserved factors. The PCA-BPNN is test with stock 600008, and the results showed that PCA-BPNN can improve the prediction accuracy and is a high efficient and accurate stock prediction model.