Research On The Combination Forecasting Model Based On Iowa Operator And Its Application In The Analysis Of Stock Index Futures

In 1982,after the United States launched the world 's first stock index futures—price line composite index futures contracts,developed capital markets successively launched different target stock index futures contracts.Discover the function to study.The results show that there is a strong autocorrelation between index prices,and the price of the futures market is ahead of the spot market,and the change of the stock index futures return is ahead of the corresponding target spot index return.The launch of stock index futures in China's market is more than 30 years later than in the United States.China is in the early stage of the development of stock index futures.Domestic scholars' research on the function of stock index futures price discovery focuses on the stock index futures launched in foreign developed financial markets and the trading time listed on the Chinese market.The longer Shanghai and Shenzhen 300 stock index futures,the research results indicate that the stock index futures have a certain price discovery function.However,the existing research literature rarely involves the research on the two new varieties of SSE 50 stock index futures and CSI500 stock index futures.Therefore,this paper selects the relevant data of CSI 50 stock index futures and its corresponding spot underlying assets for empirical research.At present,there are many prediction methods for stock index futures.The vast majority of investors use technical analysis based on stock index trends,graphic shapes,and popularity indicators.Due to the numerous analysis methods,they lack the theoretical support of scientific systems,and the independence of each indicator is relatively high.Strong,so the prediction accuracy is not high.Artificial intelligence is an integrated science,covering computers,psychology,image processing and other knowledge.In recent years,it has made breakthrough progress and wide applications.Neural networks are a branch of artificial intelligence.Wavelet neural networks are based on traditional networks.Above,introducing wavelet transform to transform it has the characteristics of nonlinear approximation of neural network,self-organizing learning,simple structure,etc.At the same time,it also has the black box identification ability of wavelet analysis,which can greatly enhance the prediction effect.This paper studies the prediction effect of the model combining wavelet and neural network in different forms,and combines the improved model with combined prediction toimprove the prediction accuracy.In this paper,the neural network model chooses BP neural network,and the wavelet analysis combines the BP neural network with separate and embedded methods to form a prediction model,and uses the SSE 50 stock index for empirical analysis and comparison.In the separated wavelet neural network model,the wavelet is used to denoise first,and then the neural network is used to train and learn the denoising signal.The way to improve the model in this model is to improve Morlet wavelet to B-spline wavelet and Daubechies wavelet to multiwavelet when denoising with wavelet,thereby improving the denoising effect and thus the prediction accuracy.Compared with Morlet wavelet,B-spline wavelet is continuous and symmetrical,has many good properties,and has a wide range of applications in real life;multiwavelet contrast Daubechies wavelet can be transformed in multiple dimensions through multiple wavelets It can be dug out.It combines with the neural network to make the training more full,the accuracy is higher,and the network structure is relatively stable.In the embedded wavelet neural network model,this paper improves the traditional single hidden layer neural network model into a double hidden layer neural network model,and combines with multiple wavelet functions to construct a single hidden layer and double hidden layer multiwavelet network model.This model structure changes the learning mode of the wavelet neural network.The double hidden layer multi-wavelet neural network has two hidden layers.Compared with the single hidden layer,its learning performance is stronger,and its prediction accuracy is higher.All the advantages of the previous wavelet neural network models are retained to improve the prediction accuracy.At the same time,this paper also introduces the induced ordered weighted average IOWA operator,and uses the characteristics of this operator to construct the combined prediction model.In fact,as far as the same single-item prediction method is concerned,its performance at different times may be different,that is,the prediction accuracy is higher at a certain point in time,and the prediction accuracy is lower at another point in time.Therefore,the existing combination forecasting method has a defect that is inconsistent with reality.Therefore,in this paper,the induced ordered weighted average IOWA operator is introduced,and the weight of the accuracy of the fitting at each time point in the sample interval is weighted by each single prediction method,and a new combined prediction model is established based on the sum of squared errors.A method to determine the weights of the combined prediction model of IOWA operator is given.In this paper,the above wavelet neural network prediction model is empirically analyzed through the SSE 50 index.Amongthem,the SSE 50 stock index selects 256 trading days of data,using the opening price,closing price,minimum price,trading volume,and transaction amount of the SSE 50 index The indicator is used as an input for analysis,and the closing price is used as an output for prediction.The improved multiple wavelet neural network prediction models are combined to make predictions based on the IOWA operator combination prediction model,which further improves the prediction accuracy.The proposed model has played a very important application research value in the stock market prediction research.