Analysis Of Public Transport Users' Travel Behavior Based On Data Mining

In recent years,mobile Internet technology,location-based services,intelligent transportation and other areas develop rapidly.As the same time,the rapid development of the country's economy is accompanied by private cars' number increasing,deterioration in air quality and haze generation.The government and relevant departments to vigorously promote public transport mode of travel.Bus,subway and shared bike like the Mobike,more and more users choose these public transport modes of travel.An increasing number of public transport users generate a lot of travel information and data.By exploring and analyzing the user travel behavior,extracting user travel characteristics,classifying the users,help us find the law and understand the behavior of the group.And then provide theoretical support for labeling the different user groups,accurate car body advertising for the businessmen,user travel recommendation,public transport company operation management and passenger distribution.In the context of large data and intelligent traffic,in this paper,the behavioral model mining model is constructed by taking behavior analysis and feature extraction of public transport users.The purpose is to analyze the characteristics of user travel at the same time mining different types of users of the potential behavior patterns.This paper starts from the research background and significance of the analysis of public transport users,the research situation at home and abroad,based on the Hadoop large data platform,do the data preprocessing of multi-source public transportation data,Combined with the network data and POI data divide the city to small regional division and attribute analysis.Using the data analysis and mining methods to extract the travel frequency,travel mode,time dimension,spatial dimension and other travel characteristics of public transport users.Through the calculation and analysis of the user's trajectory,propose the hierarchical clustering algorithm based on grouping.Combining large data processing engine Spark to improve hierarchical clustering based on distributed quadratic algorithm improvement.In the calculation of efficiency and speed of operation has improved.On the basis of the above work,this paper presents the identification algorithm of Shenzhen IC card based on SVM.Select the public security departments to confirm the thief and ordinary Shenzhen users as a positive and negative samples.Constructing 12-dimensional eigenvector matrix using proposed traveling characteristics,using support vector machine to train vector to classify thieves and ordinary users.At the same time based on the hierarchical clustering algorithm for the same line users,to speculate the thieves associates IC card.Classification results are good,and the correct rate achieves a satisfactory result.The research results of this paper have been applied in Shenzhen passenger flow analysis system and public security Shenzhen through IC card analysis and identification system and other projects.