Research And Forecast Of Stock Index Futures Based On Hybrid Model

Futures market is an important part of financial market in China,contains the regularity and characteristics of economic development,the trend of the futures and wave can reflect the status of the many aspects of a country,investors want to can have a way to predict the futures price volatility roughly,it can help investors make better investment decisions.For the analysis of futures,we can study the previous data and use time series analysis and other methods to mine the data for analysis and prediction.At present,the research on futures market can be divided into two categories:one is fundamental analysis method,which is highly subjective and more dependent on human judgment;the other is mathematical model construction based on historical data,which pays more attention to objectivity.This paper chooses the latter for analysis.Although mathematical model-based forecasting methods have a number of preconditions,these are often made from the perspective of a single model.However,the use of a single prediction model can only study the rule of data from a certain aspect,which has limitations.If multiple prediction methods can be combined to complement each other and make the best use of existing information,better prediction results can be obtained.In this paper,the ARIMA model with strong linear fitting ability is first used to model and predict the stock index futures data through preprocessing,stationary test and processing,model ordering,parameter estimation and other processes.Then considering the neural network has strong nonlinear learning ability,neural network and cycle length of the memory depth(LSTM)neural network for data before have memory function,and use the LSTM depth of learning model,this paper through normalization,set up the network structure,parameter adjustment and other network model to achieve the optimal,the change law of mining futures can make it better.To take advantage of the two kinds of prediction model,improve the quality of prediction,we in the two kinds of prediction model is established on the basis of the hybrid forecasting model,this paper adopted based on the error correction method and based on the optimal weighting method to build a portfolio model,through the RMSE and MAPE evaluation index to measure the precision of the model test,found that the two kinds of hybrid model's fitting precision is higher than that of a single model fitting accuracy and eventually adopted a hybrid model to predict financial time series data.This paper analyzes the high-frequency data of domestic csi 300 stock index futures with an example,conducts model training fitting and analysis on it with the help of Python,and verifies that the fitting accuracy of the mixed model is higher than that of a single model.Finally,we build a mixed model for the stock index futures data,and conduct fitting and prediction on it.Good results were obtained.