The Research Of Combination Model In Stock Price Forecasting

China is develeping toward the international advanced financial countries.As an important part of the socialist market economy,the stock market can help investors make reasonable and effective investments through the prediction of stocks.At the same time,effective predictions can also bring more profits to individual investors or institutions,and they can make more contributions to China's economic development.At present,investors,experts,and scholars roughly divide the prediction of stock prices into three categories: traditional investment analysis methods,modern statistics methods,and artificial intelligence methods.The traditional method is mainly analyzed from the aspects of macroeconomic indicators,national policies,industry development,financial status of the company,and stock charts and related indicators.From the perspective of modern statistics,the stock price sequence belongs to the time series category,specially the ARMA model,ES model,Markov model,etc.are widely used in stock price forecasting.In artificial intelligence methods,support vector machines,genetic algorithms,neural networks,and fuzzy analysis are gradually introduced into stock price prediction,and attempts are made to improve the optimization algorithm.Although the current research on stock price forecasting has been very mature,the analysis of stock price from multiple perspectives and multiple methods is still the goal pursued by many scholars.Therefore,this article uses a single model and a combination model to empirically analyze the time series of the closing date of China Life's stock from August 1,2016 to January 31,2018.First,the ARIMA model and GARCH model in the statistical analysis method and the BP neural network model in the artificial intelligence algorithm are used to fit a single model.Secondly,according to the idea of optimal weighting,the weighted coefficients are determined by the reciprocal method of variance and the least squared error sum rule.Three combined models are used to form a combined forecasting model.Finally,the prediction accuracy of the model is evaluated by the criteria such as average absolute error,mean square error,and root mean square error.The experimental results show that the prediction of the combinatorial model established by the combination of ideas is better than that of the single model.