Research On LOGIT Mode-split Model Based On BP Neural Network

Traffic Mode-split Prediction is one of important phases of four-stage method of traffic planning theory. The current study methods are mainly aggregate analysis method and disaggregate analysis method. The latter considers a single traveler as an analysis object and makes full use of the data of each sample and obtains the probability value which describes individual's choice behavior. It is one of the current focuses in the field of traffic planning.Logit model is one of disaggregate models applied widely on the research of traffic mode-split. It is a discrete choice model based on RUM(Random utility maximization). The key to solve this model is to obtain the utility value. The model considers the value as a linear function of personal characteristic and choosing limb characteristic. But in fact the value is customarily a nonlinear function.The thesis puts forward Logit model based on BP neural network which improves the Algorithm of the utility value using neural network nonlinear approximation ability and Matlab-ANN Toolbox. By substitution of the data included the dimensionless attribute values of influence factors and the actual selecting result of personal trip as input and output into BP ANN to study and train, the method can obtain the weight coefficient of each influence factors, thus calculating the utility values of each traffic modes. Then, the probability of each traffic modes can be determined.Finally, the thesis applies the Logit model based on BP ANN to a calculating example. The result shows that this method has good accuracy and more realistic.