A Customer Fraud Detection Model Study Based On Data Mining In Telecommunication Business

Today, reforming with corresponding by the communication technical development and foundation facilities developments perfect along with the reorganization of the telecommunication industry, each telecommunication company launched the vigorous competition among business and customer realms. The operating model is gradually transferring from technology-driven to market-driven and customer-driven, that demands telecom operators make the strategy of regarding the customers as the center. So customer's resources are becoming the competing focus of business enterprises. In customer's resources competition war, each telecommunication company's market competition resulted into lower net threshold, which gave the customer the more benefits to attract customer for the purpose of increasing the share of market. In all telecommunications customer, although the malice owes occupied only very small part of customer's total community, the loss of the group that customer resulted in the telecommunication industry is enormous, each telecommunication company in order to avoid, save or alleviate this part of losses have to adopt every kind of measure.The purpose of this paper is to research and implementation a customer fraud detection model in telecommunication. It must have better accuracy and effectiveness. Theories of data mining and relative arithmetic are introduced .Then on the basis of a actual project, the design and implementation of fraud detection system are realized according to the CRISP-DM framework. The sequence of demonstration is business understanding, data understanding, data preparation, modeling, evaluation and development.During the model construction of the study, the Decision Tree approach was adopted. Firstly, training subset with a desired class ratio was created according to the cheat-customers' history telecom data. Subsequently, based on the training subset, the detection model was established by using C5.0 algorithm. Finally, an overall prediction model was validated by using the test dataset. The result demonstrated that the model has high prediction accuracy. The main contribution of this study is to provide a highly efficient fraud detection model constructed by decision tree method. And meanwhile, it also provides an efficient and feasible technology method to solve the problem of customer fraud for most Chinese telecommunication companies that cannot get satisfied solutions through administrative methods.