Research On Recognition And Response Model Of Fraud Phone Based On Bill Big Data

At present,China's telecommunications network fraud has been in a high incidence for a long time,and the safety of people's property is facing a serious threat.Telephone fraud is the most common and most serious way of fraud.However,due to the tightness of the telephone fraud,the concealment of the means and the long chain of crimes involved,it is difficult to fight and govern,and the post attack on the telephone fraud has not reached the goal of clearing up.How to discover telephone fraud,prevent the occurrence of telephone fraud and minimize the harm to the people and society has become a very important research topic.In order to solve the problems faced by anti telecom network fraud,this paper constructs the model of fraud phone recognition and response to identifiy the fraud phone in a specific time which provides effective technical support for the anti telecom network fraud platform.Firstly the paper accomplished the task of data preprocessing:It builds the architecture of data ETL which saves the MC bills data from the business database to the data warehouse of the anti telecom network fraud center.Based on principal component analysis,the attribute features of the single data are extracted.Classifying the bill data and adding feature fields also belong to the scope of research.Secondly the paper proposes a model of fraud phone recognition: After extracting features of number's activity,features of the call behavior,features of social network,features of pairing,features of the region and features of the virtual pairing,the paper has build the flow of behavioral events for fraud phone based on the bill data of telephone fraud faking in public security institutions.The fraud phone recognition model is trained off-line based on the random forest algorithm,and the performance of the model is improved by the tuning test.The paper presents a methodology for detecting fraudulent telephone online: After statisticing the behavior characteristics of suspicious numbers and determining the credit of the number and loading the fraud phone recognition model,the paper establishes the logic of public analysis,number analysis and getting records analysis which could detect the fraud phone online based on the Spark Streaming.The parameters of the model are adjusted to meet the data throughput requirements.Finally the paper proposes a model of fraud phone response based on multi party linkage which providing a new solution for anti telecom network fraud.