| With the development of social economy,an increasing number of government projects have been organized in the form of public bidding,the supervision workload is growing up simultaneously.Before the popularization of Big Data and Knowledge Graph technology,the investigating methods of supervision department are usually passively receiving anonymous reports from the crowd,which covers very small amount of all the cases.Meanwhile,means of bribery become more and more imperceptible and complicated,increasing the difficulty of collecting evidence.This thesis designs and implements a bidding early warning system based on knowledge graph technology,it can provide quantized and explainable risk analysis,to help supervisors capture the possible risks in bidding notices quickly,and locate the company of high risks.The system is designed modularly and has four modules as below:1)Information acquisition module:focusing on acquiring the meta data.This module crawls notices of bidding and enterprise information from different website,it contains the configuration of the crawlers and captcha dealing;2)Knowledge extraction module:focusing on extracting particular information from unstructured and semi-structured data.This module works in the idea of "Divide and Conquer",it classifies the meta data according to administration division,into several structure-similar classes,then extracting the target sentences one by one;3)Knowledge merging and storage module:focusing on integration and storage of structured information.This module uses an open-sourced Graph DBMS called Neo4j,it imports the knowledge data after merging all the entities,providing extra multi-degree query and visual interface;4)Knowledge application module:focusing on early warning of risks.The early warning consists of qualitative analysis and quantitative calculation,qualitative analysis is to analyze the public opinion of the company and see the risks,quantitative calculation is scoring the risk level of a company by its industrial and commercial properties,both parts can trigger the alarm.The system has been deployed and been working stably,helping supervision department filter the bidding project of high risks,locate the source of risk,finally elevating the efficiency. |
