Stock Price Predicting Based On The Composite Model Of Improved GM(1,N) And Optimized SVM

The stock market is an important part of the capital market, which is also seen asan economic "barometer", playing an important role in optimizing capital allocation,financing and increasing the value of assets and other areas. However, consideringthe coexistence of the stock market risk and return, high investment risk, using somemethods and related models to analyse financial market trends, estimate and predictthe stock price, providing investors with investment decision reference, has a veryimportant practical significance.By reviewing and summarizing the literatures,this paper analysed the factorsaffecting stock prices from two angles of macroscopic and microscopic,initiallyselected GDP, interest rates, exchange rates etc eight indicators as input variables ofstock price forecasting.In order to streamline the number of input variables toimprove the prediction accuracy, we used the improved gray relationalanalysis,which is combined with the stepwise regression, correlation analysis,tomake the secondary feature extraction of the variables affecting the stock price, andultimately selected M2, exchange rates, business climate index, S&P500,fourvariables, as input variables of the predictive model.In stock price prediction,we constructed combined model based on inducedordered weighted averaging operator, combined model covered improved GM (1, N)model and support vector machine model optimized by uncertain knowledge PSO(PSO-UK), two models are used to forecast the linear and nonlinear parts of the stockprice using induced ordered weighted averaging operator to determine the weight ofthe two models.previously. we took an empirical test using identified four variablesas input values, the Shanghai Composite Index as the output value of the model.Using the mean absolute percentage error, mean square percentage error andother indicators,we compared the combined model with original GM (1,1) model,original GM (1, N) model, support vector machine model optimized by particleswarm and BP neural network model, the results showed that the combined modelproposed in this paper has better generalization and predictive ability,verifying thesuperiority of the combined model.Combined model can provide reference forinvestors in the stock market to make investment decisions.