A Stock Index Futures Price Prediction Scheme Based On The LSTM-SVR Hybrid Model

Stock index futures is a kind of futures whose object is stock price index.Stock index futures play the role of price discovery and risk aversion,and also enrich the investment strategies of investors.Stock index futures market has become an important part of China’s financial market,which is of great significance for the prediction of stock index futures price.The existing methods of stock index futures price prediction mainly include time series analysis,support vector machine regression model,neural network prediction method,etc.But time series analysis method is difficult to describe the randomness and complexity of financial market volatility,and the prediction effect of financial time series is not good,while support vector machine has the problems of kernel function selection,slow training speed and consumption In the first mock exam,the neural network has the disadvantage of being difficult to determine the network structure,which makes it possible for a single model to be unable to model the model completely.Therefore,this paper proposes a hybrid model construction scheme which integrates the support vector machine regression model and the long and short-term memory model in the field of neural network,in order to further improve the quasi determination of stock index futures price prediction.This paper first analyzes three kinds of stock index futures in China.In view of the speculative and easy manipulation of the 500 stock index futures in China,this paper chooses the more representative data of the main contract days of CSI 300 stock and 50 stock index futures as the research object,and studies the two futures products to avoid the contingency of the model.In the selection of indicators,the technical indicators are included in the input variables of the model,and the influence of basic market indicators and technical indicators on the closing price of stock index futures is comprehensively considered,and the principal component analysis method is used to reduce the dimension of variables.In the actual model construction,this paper extends the long-term memory model based on deep learning to the field of stock index futures price prediction,uses the latest Adam algorithm to optimize the model,and creatively combines the longterm memory model with the commonly used support vector machine regression model through the ridge regression method to form a hybrid model.The empirical results show that the hybrid model is better than the single LSTM and SVR.The model has a smaller error in the prediction of the closing price of stock index futures,and has achieved a better prediction accuracy.Finally,based on the LSTM-SVR ridge regression hybrid model,this paper makes two kinds of unilateral trading strategies of stock index futures,taking the main contracts of 300 stock index futures as an example,and the results show that no matter which strategy has achieved much higher annual returns than the benchmark yield,demonstrating the rationality of the hybrid model.