Multi Resolution Prediction Model Based on Wavelet Analysis and Neural Network

Abstract

At present, the securities market is mainly composed by the stock market. The accurate prediction of stock price means that investors can get high returns and the government can effectively supervise the market. The future trend of a country's stock market is constantly changing and influenced by many interference factors. The traditional prediction methods can not effectively solve the problem of accurate prediction of such nonlinear systems. The time series data of stock price formation are nonlinear and non-stationary, which are difficult to deal with by traditional models. Wavelet neural network has a good ability to approximate the nonlinear mapping. Therefore, a wavelet neural network prediction model is proposed to deal with non-stationary data. Wavelet transform divides time series into high-frequency and low-frequency sequences. According to the data characteristics of different sequences, a prediction model is established. The experimental results indicate the lower prediction error of the prediction model after comparing it with the improved neural network and GARCH method.