Design And Implementation Of Telecom Fraud Identification System Based On Value Iteration

With the advent of the era of big data,people's privacy information has been leaked more and more seriously and telecom fraud is also intensifying.A communication record will be generated to record the original communication information when people make a call.Traditional machine learning methods extract the statistical information of communication records to identify fraud users.But it cannot fully mine the associated information by only taking advantage of users' short-term communication behavior.In order to make full use of the associated information contained in the communication records to identify fraud users,this paper extracts personal feature vectors and intimacy between users from users' communication records to construct communication graph representing users'communication behaviors.Since the communication graph is a non-European spatial structure,the paper uses a graph convolutional neural network to predict users' local credibility.Users' local credibility is used to measure the short-term credibility,and global credibility is used to measure the long-term credibility.In addition,under the limitation of not affecting the real-time performance of the system,the paper introduces iteration into the updation process of intimacy and global credibility,and designs and implements respectively the iterative updation algorithm of intimacy and the iterative updation algorithm of global credibility.Based on the communication graph model and algorithms,this paper designs and implements a telecom fraud identification system based on value iteration.Firstly,the thesis gradually decomposes the functional requirements of the system and summarizes the non-functional requirements.Secondly,this paper introduces the graph convolutional neural network layer for telecom fraud,the local credibility prediction model,the iterative updation algorithm for intimacy and the iterative updation algorithm for global credibility,which are verified by experiments.After that,the paper designs and implements the system in detail around the functional requirements of the system,and completes the functional tests to verify the usability of the system.Finally,the key content of the paper is summarized and future directions are pointed out.