Research And Application Of Time Series Forecasting Model Based On Conjugate Gradient Method And Optimization Theory

Forecast is based on the known information,the establishment of the system model,according to a method of the unknown things to be measured in order to control the development of things have control.With the development of society and the progress of science and technology,the forecasting method has also been developed on this basis.In practice,the time series usually show a dynamic structure of highly complex nonlinear state.Only by a kind of prediction method has not been able to better carry out the fitting prediction research,the different prediction methods for the right combination of research,which can be integrated with different kinds of forecasting method advantages.In this paper,a modified method of prediction method is proposed,which can improve the fitting precision of the model.The result proves that the effect is better than that of using a certain prediction method,and it has been widely used and recognized.The results show that the proposed method is better than other methods by introducing the conjugate gradient method to optimize the original single model.Firstly,a new conjugate gradient method is proposed on the basic idea of nonlinear conjugate gradient method,and the validity,practicability and its superiority of the new conjugate gradient method are proved by examples.Secondly,the new conjugate gradient method is used to optimize the parameters of the ARIMA model,and the procedure of parameter optimization algorithm is given.Finally,the optimized ARIMA model is used to fit the 2007-2016 retail sales value data.The analysis shows that the improved model is reasonable and effective,so as to improve the forecasting result.