上证指数基于SVD的组合预测模型

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.

Combination Forecasting Model of Shanghai Index Based on SVD

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,compared with the single model,this model has higher fitting and prediction accuracy.