Research On Anti Fraud Platform For Telecom Frud

With the development of communication network technology,people's requirements for the quality of telecommunication services are constantly improving,and the challenges to telecommunication services are growing.People are not satisfied with the previous"available"demand for telecommunications services,but tend to be more refined,personalized and high-quality.As one of the most common means of fraud,Telecom fr-aud directly affects the experience and service quality of telecom users.The problem of Telecom fraud will not only cause unnecessary harassment to users.Some Telecom frauds may even threaten the safety of users'personal and property.In the development of telecommunication industry,a large amount of relevant data has been accumulated,which can be used for the research of anti-fraud platform.Therefore,the National Telecommunications Department hopes to use these accumulated data to conduct a precise analysis of telecommunications fraud.And hope to be able to use these analysis results to propose an accurate classification of telecommunications fraud,so as to conduct targeted tracking,analysis and interception of telecommunications fraud.In the telecommunications sector for telecommunications anti-fraud demand,the anti-fraud platform project for telecommunications fraud was officially launched.Telecom fraud includes many aspects.In practice,it can be divided into two main types:recording fraud and voice fraud.Because of its obvious spectrum characteristics,audio fraud can be easily filtered by algorithm,while voice fraud is more difficult to identify,while voice fraud is the target of this anti-fraud system.The purpose of this proj ect is to realize a telecommunication anti-fraud platform.Through the use of telephone recording information,through the application of a variety of algorithms,including voice recognition,natural language processing,to analyze and mine telecommunications fraud from multiple perspectives.And using these analysis results to construct a system that can effectively intercept fraudulent telephones.Compared with the previous anti-fraud system,which can only intercept suspicious traffic by user active marking,this anti-fraud platform is more reasonable.Direct use of user call information to intercept,there will be no previous anti?fraud system"can not identify the number change" problem,can better intercept the relevant telephone when the fraud occurs or even before it occurs.In order to realize this system,this paper first consulted and studied the technical data in related fields to ensure the technical feasibility of the whole system.Then,the project requirements are systematically analyzed,and the system modules are divided according to these requirements.The function of each module is designed and divided at the level of software engineering.In order to make the result of speech recognition more accurate and more convenient for analysis,this paper proposes a novel model based on deep peech 2 architecture.Compared with deepspeech2,it can make better use of a variety of voice information to achieve better results,and for the sub-problem of voice classification of customer service telephone.This paper presents a speech classification algorithm based on I-CNN,which can better adapt to the characteristics of customer service telephone.For the results of speech recognition,a novel text classification method is proposed in this paper.In view of the inaccuracy of speech recognition results,the text of speech recognition is not directly used as input,but the probability distribution of speech recognition results and recognition results are simultaneously fed into the text classification model to improve the effect.These key algorithms have been validated by many experiments designed in this paper,and have been approved by professional academic journals and conferences.These key problems solved in this paper make the design and implementation of the whole system possible.In this paper,around the key issues,requirements and detailed design of the system,a set of reasonable and available anti-fraud platform for telecommunications is constructed.After testing in this paper,the platform can well meet the demand of telecommunication anti-fraud.At the end of this paper,the realization of the system is summarized and prospected.