上证指数基于SVD的组合预测模型  

Combination Forecasting Model of Shanghai Index Based on SVD

作　　者：刘常明[1] 张德生[1] 李金凤[1] 任世远[1] 

机构地区：[1]西安理工大学理学院,陕西西安710054

出　　处：《陕西科技大学学报（自然科学版）》2012年第2期122-125,139,共5页Journal of Shaanxi University of Science & Technology

摘　　要：股票指数时间序列具有非平稳和高噪声等特点,在进行股票指数预测时,由于噪声的影响,单一模型的预测精度往往不高.作者建立了基于奇异值分解(SVD)的BP神经网络和ARMA-GARCH组合预测模型,该模型将原序列分解为趋势部分和噪声部分,分别进行研究.实证研究结果表明:该模型的拟合、预测精度较高.Stock index time series with non-stationary, high noise character, making stock index forecast, due to noise, the predictive accuracy of a single model is often not high. In this paper, the BP neural network and ARMA-GARCH combination forecasting model based on singular value decomposition（SVD） is built, the original sequence is decomposed into tend and noise parts and studied respectively. Empirical research result shows that, com- pared with the single model, this model has higher fitting and prediction accuracy.

关 键 词：奇异值分解 BP神经网络 ARMA-GARCH模型 

分 类 号：O211.64[理学—概率论与数理统计]