Research On Frequent Pattern Mining Of Taxi Passengers Based On Distributed Platform

In recent years,the number of urban residents is increasing,the scope of residents’ activities is expanding,and traffic congestion has become a persistent disease.More and more scholars solve urban traffic problems by studying frequent modes of passenger travel.The traditional research methods are usually questionnaire survey,which leads to the failure to find the rules of passenger travel activities accurately.At present,most taxis in cities are equipped with GPS monitoring system,which is mainly aimed at monitoring the operation of taxis,and can generate a large amount of data.These data objectively record the travel status of taxi passengers and the busy conditions of urban roads.With the development of the whole human society information,the number of taxi passengers data has increased.Collecting these data can tap the potential value of taxi passengers’ travel.This paper studies the frequent travel patterns of residents in different time periods according to the spatial and temporal distribution characteristics of taxi data.Firstly,to solve the problem of adding travel path,the functional areas of taxi data,micro blog data and POI data are used to identify the area by map segmentation technology,so as to give semantic information to location points.Then,in view of the problem that the frequent subgraph mining algorithm in the frequent pattern mining algorithm can not be applied to the existing weight graph model,this paper improves the graph model.In addition,for the existing graph model,muledge algorithm mining is proposed to mining frequent association patterns of passengers’ travel.Finally,in view of the problem that single machine can not process and mine data quickly,this paper builds hive distributed database and spark computing engine platform to mining frequent association mode of passenger travel.The database design of taxi GPS achieves the effect of query optimization.Two distributed mining algorithms are set up: the existing fsmbus algorithm and the parallel spark algorithm of muledge algorithm＿Muledge algorithm.In this paper,the spark cluster platform is built by 5 workstations to store and mine taxi GPS data.Urban road network data,POI data,taxi GPS data and micro blog sign in data are used as experimental data of mmorftp method.The experimental results show that mmorftp is effective and feasible,and the frequent travel patterns found by relevant research can provide decision-making basis for road planning,traffic management,commercial layout and other applications.