The Research On Combined Prediction Of Improved Grey Neural Network And Application

In real life there are many complex time series,for traditional forecasting methods, the forecasting precision often can't meet the need because of its own limitations and the excessive disturbance factors ,since improved grey predition model and nerve–network prediction have complementary strengths ,an algorithm for combining the two together can be greatly improved forecasting accuracy.This paper,based on the process of establishing several improved equal- dimension-newly-information grey models ,find that the superior forecasting value is not necessarily equal-dimension–newly- information value, but the combination of the forecasting values calculated by the above several improved GM(1,1) models. Therefore,BP network are used to seek the function of several predictive value and the actual value and to get the best predictive value.The algorithm based on equal-dimension–newly -information grey models and grey BP network are presented ,the time sequence of National Private automobile volume test the validity of the algorithm.In this paper,it is shown that the data should be revised and processed properly in advance according to the fact rather than be directly used to set up the model when the system disturbance is too great or system behavior mutates instantly and coming out serious disturbance system abnormal data.Meanwhile, the improved grey network combined forecasting model based on the revised data is presented. According to the time sequence of passenger transmission volume in Nanchang Railway Station,several models are set up ,From comparison of model predictions step by step,it is indicated that revising date,improved grey and improved grey network,improved grey network combination can really improve prediction accuracy. The algorithm provides a feasible method for this kind of time sequence forecast.