Research On Path Selection Optimization Model And Algorithm Based On Difference Analysis

With the rapid socio-economic development and urbanization,motor vehicle ownership has been increasing,leading to many traffic problems,which greatly affect economic development.An effective way to alleviate urban traffic problems is to study the characteristics of travelers’ path selection behavior,which can provide technical support for urban traffic planning.Most scholars’ research is based on some assumptions to analyze the traveler’s route choice behavior,but this study takes the traveler’s GPS track data as the basis to reflect the traveler’s route choice behavior on the electronic map,which makes the research results of the traveler’s route choice behavior closer to the reality and also provides a more scientific and reasonable theoretical basis for the traffic management department to formulate the traffic policy.The study of urban traffic travel behavior is the core problem in the field of urban transportation research.In this paper,the theories and methods related to data processing,path selection set,neural network model and decision tree model are firstly explained,and then the personal information,family information,work information and their GPS track data of travelers are collected by means of questionnaire survey with urban residents in Shenzhen as the research group.Based on the above data,the key factors affecting the deviation of the actual path from the theoretical path were clarified by analyzing the factors affecting the travel path choice of travelers;considering the differences between the actual path and the theoretical path and the factors influencing the differences,the path choice set was established by using the link penalty method,and then the commuter travel path choice model was constructed by using PS-Logit as the carrier,and at the same time,for the study of multi-day single At the same time,in order to study the path selection behavior of travelers in multi-day single OD,different from the previous path selection models,this paper establishes a neural network model and a decision tree model to classify travelers separately,and then predicts the path selection behavior of travelers,and compares and analyzes the prediction effect of the two models,and finally performs sensitivity analysis on the neural network model and the decision tree model respectively,and analyzes the effect of the number of neurons in the hidden layer and the learning rate on the The effects of the number of neurons in the hidden layer and the learning rate on the performance of the neural network model and the number of leaf nodes on the performance of the decision tree model were analyzed.The analysis results show that(1)there were large differences between the actual travel path and the theoretical path in terms of overlap,time difference and distance difference,etc.The longer the distance of the actual travel path,and the overlap ratio was between 0-20%,the larger the deviation value of the distance difference between the actual travel path and the shortest distance path.And the factors that have significant positive utility effects on the distance ratio are age and the detour of the actual path;(2)When the actual travel distance of commuter travelers is within 3 km,the time difference between the actual travel path and the shortest distance path is almost unchanged,while the deviation value of the time difference is the largest when the number of intersections is greater than 1.5.The factors that have a significant positive utility effect on the time ratio are the working time flexibility,the straight-line distance from the origin and destination,and the number of intersections per kilometer of the actual path.(3)The calibration of the parameters of the PS-Logit model found that the percentage of expressways had a positive utility effect on the travel path choice of travelers,while the path length and detour degree had a negative utility effect on the travel path choice of travelers.(4)Classification of travelers by multi-day single OD For traveler classification,the accuracy of neural network prediction was about 66%,while the accuracy of decision tree prediction was 58%,and the results of neural network prediction errors were relatively more concentrated than those of decision tree model prediction,while for the prediction of traveler path choice behavior,the accuracy of neural network model prediction was 67% in the prediction results of path choice behavior The accuracy of decision tree model prediction was 54%.The results of neural network prediction outperformed the prediction results of decision tree model,and both obtained better prediction results.