The Research Of Data Mining In Fraud Management In Mobile Communications

Fraud brings a huge loss to the mobile communications industry each year. The third generation mobile communication services will face even greater threats and challenges of fraud due to its own characteristics. In order to maintain the profits of majority of mobile operators and customers, to improve the communication quality of service, it is necessary for mobile communication operators to analysis the various types of existing fraud, to find out the root causes of fraud, and then take effective preventive measures to reduce the mobile operators'loss caused by fraud.Data mining technology is a rising information technology of subject-oriented analysis and decision-making management supporting in the last two decades. This technology is undoubtedly a good solution to solve the mobile communications fraud issue. The application of data mining technology can not only take advantage of the mobile operators'mass of data resources but also help business decision-makers from the traditional reliance on historical experience and intuition to judge the behavior of fraud into the use of informative and intelligent technology to detect the users'fraud behavior patterns, greatly reducing the company's operating costs and improving the forecast accuracy.In this paper, we collected a large number of mobile call detail records and user information data from the data warehouse, and used three types of classification algorithm (decision tree, Bayesian and neural network) provided by data mining tool WEKA to create a fraud prediction model respectively. We did a comprehensive assessment to choose the most appropriate algorithm model for this research field by comparing their prediction accuracy, false positive rate and running speed, and other aspects. Next, we made use of the research result of previous stage to design a fraud management system that can prevent, detect and block the fraud according to business needs. The system can help mobile communications operators to do a better job of fraud management.