| Along with the telecommunication industry business is increasing rapidly and the income of the operator is rising steadily, telecom fraud is becoming more and more seri-ous. These fraud behaviors not only affect the further development of the telecommuni-cation business, but also bring huge economic losses to the operator, increase the opera-tor’s network load, lower customer’s communication quality. The traditional anti-fraud technology is mainly based on anti-fraud rules, the setting of the rules depends on the business personnel’s experience, and is easy to cause the misleading and leakage judg-ment. In this paper, we adopt the new data mining technology, analyze, modeling the classic SIMBOX fraud, and give a complete implementation schema.In this paper, we first introduce the concept of data mining, tools and key algorithm, then the SIMBOX fraud is analyzed, finally we follow the steps of requirement analysis, design, system implementation to describe the whole development process of the system. The system is based on the popular B/S architecture, using Oracle as the database,Java as the programming language, Clementine as the data mining modeling tool.The system main functions are as following:Data Source Configuration, Index Co-nfiguration, Mining Modeling, Runtime Task Configuration, Data Collecting and Pre-processing, Index Statistics, Fraud Data Mining and Fraud Monitoring and Processing.This system analyzes the user’s history call data to dig out the fraud model. As a contrast with the traditional methods, the new method lower the requirement of the user skill, improves the anti-fraud accuracy, reduces misleading rate and leakage rate. But because the user history call data has the certain lag, it is the future research direction and focus that how to combine the real-time call data and signaling data to analyze the fraud and improve timeliness of anti-fraud results. |
