Analysis Of Metro Passenger Travel Law And Study Of Destination Prediction Method

With the rapid development of China's urban rail transit,many cities have completed and opened urban rail transit lines.At the same time,Automatic Fare Collection system(AFC)has been widely used in China's urban rail transit system,which makes the collection of basic passenger flow data for urban rail transit more complete and accurate.How to collect data based on the passenger flow of the AFC system,and use data mining technology and machine learning mechanisms to achieve more accurate passenger flow prediction of the urban rail transit network,has important practical significance for urban rail transit operation and management.Based on the history of the passenger's subway card data,this paper frist analyzes the individual passenger travel rules,then speculates the passenger's destination,and finally predicts the OD passenger flow of the future urban rail transit network.The main work of this article is:(1)This article gives a brief description of the data source and AFC system,summarizes the reasons that may cause abnormal AFC data,summarizes the types of abnormal data,proposes data cleaning rules,and cleans the data.Then use AFC data to analyze the space-time law of subway passenger flow.(2)Using the ID attribute in the card data,the historical card records of individual passengers within a month were extracted,and the individual passenger travel rules were analyzed and studied.According to the characteristics of AFC credit card data,subway passengers are divided into two types: regular trips and random trips.This paper proposes a density-based clustering(DBSCAN)method to perform passenger cluster analysis.Based on the clustering results,the passenger travel time-space law is described and analyzed.(3)Based on the travel laws of individual passengers,this paper calculate the ratio of the total number of trips per passenger to the total number of trip ODs,and the ratio of the number of random trips to the total number of trips.Further,the two-step clustering method was used to classify rail transit passengers into three categories: commuters,random passengers,and flexible commuter passengers.Based on the travel characteristics of different types of passengers,this paper presents a method for predicting the travel destination of passengers after entering the station(4)On the basis of the above research methods,the specific individual passenger destination prediction method is extended to the entire network structure of rail transit to realize the prediction of the OD flow direction of the entire network at the same time.Finally,using the whole network card inbound information of a certain day,based on the python programming language and Oracle database,the outbound probability of different sites,the passenger flow of different OD pairs,and the OD distribution of the entire network were predicted,and the prediction results are analyzed and analyzed.