| With the rapid development of Internet plus,more and more industries are increasingly integrated with the Internet.Among them,freight companies represented by freight pulling,Didi Freight and Kuaigou Taxi are the most prominent.However,with the continuous development of business,black products have become more irregular in the Internet field,especially in the transaction and payment scenarios.Today,although most risk control business scenarios have mature solutions,querying all relationships and behavior associations between users in one interface is time-consuming,labor-intensive and inefficient.Finding users who can be directly related also requires experience accumulation.When new users join,they also need to re accumulate experience.In the face of the problems caused by the above scenarios,graph databases have been generated and developed for a long time.As the first open source graph database in China and included in the Apache incubation program,Huge Graph has the characteristics of good stand-alone performance and supporting distributed storage,and has become the first choice of all companies.This paper first investigates the graph database and graph computing engine,and analyzes the requirements of the graph application platform,mainly including the demand analysis of graph data sources,the storage demand analysis of graph data,the demand analysis of graph algorithms,and the demand analysis of graph visualization display.Finally,we chose to implement a graph application platform based on the graph database Huge Graph and the graph computing engine Spark Graphx.Figure The overall design of the application platform is divided into basic components,intermediate components and application components.Basic components include data import,data storage and data management;Intermediate components include data request and graph calculation;Application components include data visualization and result display.Graph application platform can be divided into storage module,algorithm module,service module and visualization module.In the storage module,a data importer is designed to solve the performance problems when data is imported into the graph database;In the algorithm module,Page Rank,Triangle Count,SSSP and other graph algorithms are implemented;In the service module,a service platform based on Spring Boot is designed,through which users can query and analyze data;In the visualization module,the chart data is visualized through the Echarts framework.Finally,the graph application platform is applied to the freight risk control direction.Based on the design,Huge Graph Spark implements Louvain,improved LPA and other community division algorithms,abstracts the complex relationships in black production data into mathematical expressions,obtains hierarchical community discovery results,and identifies gang fraud.In the large-scale complex social network with massive data,the distributed computing framework Spark Graphx further improves the operation efficiency of the algorithm,while ensuring the accuracy of the algorithm. |
