Research On The Construction And Mining Of Passenger Knowledge Graph In Urban Public Transport System

With the high-speed development of the domestic economy,first-tier cities have established a modern urban public transportation system with "subway as the backbone,bus as the extension and cab(online taxi)as the supplement".Public transportation runs at high speed every day,serving millions of passengers.As the scale of the road network continues to expand and the number of urban residents continues to increase,the density of traffic and pedestrian flow continues to grow,bringing unprecedented pressure on the relevant management departments.The traditional information technology infrastructure can no longer satisfy the needs of large-scale urban transit scenarios of operation and safety management.How to use existing data resources and big data technology to support the refined operation of public transportation and improve the safety protection level of public transportation is the primary problem faced by practitioners.Based on the massive multi-source heterogeneous data generated by the subway system of a first-tier city,this study uses relevant big data tools to construct a public transport system passenger knowledge map to support public transport safety management.Firstly,we design the pattern layer and data layer of knowledge graph according to the existing data and application requirements.Secondly,the big data computing engine Spark and related algorithms are used to calculate the spatiotemporal features,social features and electronic features of passengers to realize knowledge extraction.The constructed knowledge graph is stored in the graph database Neo4j;finally,the method is verified on a large-scale real data set.The experimental results show that it takes an average of 0.042 seconds to query the shortest path between two nodes in a graph database with 5.22 million nodes and 26.82 million relationships.The knowledge graph can effectively support the application query in the public transportation safety scenario,thus providing technical assistance for safety prevention and control in the public transportation field.At the same time,based on the constructed knowledge graph,this study uses four typical community discovery algorithms to divide the passenger group into communities,evaluates these algorithms with multi-dimensional indicators,and analyzes the results of community division,so as to effectively reduce the cost of frontend exploration,intelligence investigation and human resources in rail transit policing activities.