Prediction Of Passenger Traffic Volume Based On Grnn And Grey Theory

Accurate consequence of passenger traffic volume prediction is one of major basises for traffic planning. This paper aims to researching on the characteristics of prediction, the paper finds out the way to increase accuracy of prediction and studys on Grey-General Regression Neural Network model under the condition which is lack of data sample.This paper firstly summarizes the characteristic of passenger traffic volume as nonlinear and small sample through deeply researching on it. What's more, the article also gives out the methods and steps of traffic planning and analyzes the main factors which influence passenger volume.Secondly the paper analyzes the passenger traffic volume prediction methods on present, such as Regression Analysis prediction model, Exponent Smooth model, qualitative analysis and BP Neural Network model, and presents all advantages and disadvantages of these methods.Then, the author proves that GRNN model can well satisfies the demand of passenger traffic volume prediction in comprehensive traffic system based on the analysis on the application condition of GRNN theory both at home and abroad. The paper makes generation and manipulation to the raw data, and constructs the prediction model by using the generative data. After then, the author does some research on the G-GRNN suitable for comprehensive traffic prediction, and discusses three models in G-GRNN with different structures, those are PG-GRNN,SG-GRNN and IG-GRNN. According to the conclusion above, the paper studies on the G-GRNN prediction model based on the regression analysis and G-GRNN prediction mode based on the time series analysis separately, and works out the idea and steps of how the model constructed; analyzes the key problems like pretreatment of input/output data in the process when constructs model, confirming smoothing parameter and the number of input neurons on the small sample condition and then provides effective solution.Finally, the paper takes railway passenger volume of SiChuan in 1997-2005, for instance. The calculations proves that passenger traffic volume prediction based on the G-GRNN win more accuracy no matter in the fitting of historical data but in the examination by model's extrapolation than the other three models that are Regression model, Cubic Exponent Smooth model and BP Neural Network model.