摘要
研究股票价格准确预测问题,股票价格变化具有非线性、时变性,且含有噪声,单一或传统线性预测模型不能全面反映其变化规律,预测精度低,误差大。为了提高股票价格预测精度,提出一种组合的股票价格预测模型(CAR-BPNN)。首先采用主成分分析对股票价格数据进行预处理,消除噪声,然后采用CAR对线部分进行预测,BPNN对非线性部分进行预测。采用熵值法确定CAR和BPNN对预测结果进行组合,获得股票价格的最终预测结果。通过股票价格实际数据对CAR-BPNN进行测试,测试结果表明,CAR-BPNN充分利用两种模型的优点,比单一模型的预测精度更高,可以为股票价格精确预测提供依据。
Stock price prediction is current research hotspot. The stock price changes are of nonlinear and dynamicity, it is a kind of complex time series data, therefore, single prediction model can not reflect the variation, and the prediction accuracy is low. In order to improve the prediction accuracy of stock price, a stock price prediction model was put forward based on CAR and BP neural network （CAR-BPNN）. CAR-BPNN used the principal compo- nents analysis to pretreat the stock price data and eliminate all noise effects. Then the stock prices was respectively predict by CAR and BPNN. Finally, entropy method was used to weight the linear results and nonlinear results, and the prediction result is obtained. CAR-BPNN was tested by the stock price data, and the results show that CARBPNN prediction accuracy is higher than single prediction model. It can make full use of the advantages of two mod- els, and is more suitable for complicated stock price prediction.
出处
《计算机仿真》
CSCD
北大核心
2012年第1期348-351,共4页
Computer Simulation
