Research On The OD Estimation Method Based On Artificial Neural Network

Traffic demand analysis is the basic work of traffic planning and large-scale traffic organization scheme research and analysis,and OD matrix is its most basic basis.OD estimationis a method to obtain OD matrix.When it is difficult or expensive to obtain OD matrix directly,and it is relatively simple to obtain road traffic,it is necessary to backtrack OD matrix by road traffic.Because there are many factors that affect the result of the back calculation in the actual road traffic network,the existing back calculation model is difficult to accurately express the relationship between the road section flow and OD with complex internal relationship,which leads to the difference between the result of the back calculation and the actual situation.Artificial neural network can make use of its unique structure and calculation principle,and can better express the internal relationship between input and output,so this paper explores the possibility of calculating OD by artificial neural network.Firstly,in this paper,the feasibility of the neural network to deduce od is analyzed by studying the basic theory and neural network model,and the basic idea and method framework of the neural network to deduce od is put forward;by analyzing the advantages of BP neural network in OD backstepping,the neural network to deduce od model is established;aiming at the problem that it is difficult to obtain the training samples of neural network to estimation OD,the virtual model is proposed According to the OD matrix,the traffic is allocated to obtain the section flow,which is used as the training sample;then the model is built and trained based on Python;finally,the OD estimationis carried out based on the actual section flow and the effect of the estimationis analyzed.It can be seen from the results that: the capacity of training samples should not be too small.From the results of two back extrapolations,it can be seen that when the capacity of training samples is increased,the error of training samples and back extrapolation results is significantly reduced,so the capacity of training samples should be increased as much as possible;when training models,the smaller the training error of training samples,the better;when the error of individual training samples does not meet the requirements Therefore,the sample should be able to better reflect the actual traffic conditions of the project when the sample is proposed.The conclusion above adds new contents to the modern traffic planning theory.This method can be used in practical engineering and a practical method for engineering practice,which has strong practical significance.