A Separate-Predict-Superimpose Predicting Model for Stock

摘要

The purpose of this research is to propose a more precise predicting model, the Separate-Predict-Superimpose Model, for time series, especially for the stock price and the stock risk than the established predicting method. In this model, time series are separated into three parts, including trend ingredient, periodic ingredient and random ingredient. Then the different suitable predicting methods are applying to predict different ingredients to receive accurate outcome. Ultimately, the final predicting result is superimposed by the three ingredient predicting outcome. The wavelet analysis, combination predict method, exponent smoothness method, Fourier Transform, fitting analysis and Autoregressive Moving Average (ARMA) are adopted in this model. By applying the model to predict the Shanghai Composite Index, China National Petroleum Corporation stock price and risk and comparing with other predicting method, a conclusion can be made that this model can fit various characteristic time series and achieve a more precise result.