Investigation Of A Wavelet Decomposition-based Hybrid Model For Stock Index Forecasting

As the main body of financial market,stock market in the country's economic development plays a very important role.Stock price index is the index to describe the total stock price level and volatility of the whole stock market,forecast analysis of stock price index,in terms of microcosmic,affect the investors' investment strategy,in terms of macroscopic,provide the basis for the country's macro decision-making.Therefore,many researchers at home and abroad have carried out a prediction study on the stock index,and proved that it is a more effective method to analyze the stock index from the time series point of view.At present,the research on stock index time series mainly includes traditional time series model and data mining method.In the traditional time series model,the Autoregressive Integrated Moving Average Model(ARIMA)has been one of the most widely used linear models,and the Extreme Learning Machine Model(ELM))Is a general model used in data mining for nonlinear time series prediction.In this paper,a single autoregressive integral moving average model and a limit learning machine are firstly used to predict and analyze the 180 SSE stock index,the experimental results and in-depth analysis show that for complex instability,strong randomness and other significant stock index time series,using a single model is difficult to achieve accurate the prediction,that is not very good to capture the characteristics of stock index time series,the research and construction of mixed model and algorithm of time series prediction is to enhance the quality and level of the development trend of the stock index.The time series of stock index is affected by many factors,such as complex instability and strong randomness.In order to better capture the stock index time series characteristics,give full play to the advantages of linear and nonlinear models,we will stock index time series as a composite with low and high frequency time series,and the development of wavelet analysis theory and technology for the composite time series provides effective solutions for the low frequency and high frequency decomposition component.Based on this,this paper put forward a prediction hybrid model of wavelet decomposition of the autoregressive integrated moving average and extreme learning machine based on the stock index,the model based on wavelet theory,the stock price index time series is decomposed into low frequency component and high frequency component random trend,and then use the ARIMA model to predict the trend of low frequency time series in order to capture the linear law the stock price in the high frequency random time series prediction using ELM model to capture the nonlinear law of the stock price.The prediction results obtained the final stock value of low frequency and high frequency time series sequence will get forecast.In this model,the ARIMA model is used to predict the trend of low frequency linear sequence,ELM model is used to predict the nonlinear high frequency random sequence,their model advantage into full play,thus the model for stock index time series prediction effect should be improved.The experiment showed that the prediction of the SSE 180 index using the proposed hybrid model,prediction model of mixed sample prediction root mean square error and average relative error percentage than the single model results were significantly reduced,which can improve the prediction accuracy;and the relatively recent Yuan Lei proposed by ARIMA and LSSVM(least squares support vector machine)hybrid model[25]also has greatly improved the predictive accuracy.In the verification process and model of mixed model experiments we further found that the high frequency component of random directly using the ELM prediction model has not yet reached the best results,and further exploration and promotion space.In view of this situation,based on the wavelet decomposition of the autoregressive integrated moving average model and extreme learning machine mixing model was extended to further study based on the high frequency time series using ARIMA model and ELM model,the prediction of the results of the two models again input ELM model obtained prediction results of high frequency random component finally,the prediction results of prediction results of low and high frequency time series to obtain the final composite stock index.The experimental results show that the extended hybrid model can further improve the prediction accuracy.Therefore,stock index time series of the proposed hybrid forecasting method and model has some theoretical significance for promoting the research of complex time series prediction,but also has a certain economic value and application prospect.