Payment arrearage has been a serious and urgent issue in both internal and overseas telecom industries. The emergence of Analytical Customer Relationship Management (ACRM) and Data Mining (DM) brings about technical means for handling the above issue. Based on the survey of both theories and applications of ACRM and DM, this thesis integrates them to study the classification of fraudulent customers in telecom industry. The decision-tree algorithm and its deployment in the customer classification is identified and studied in terms of the real business situations and requirements in mobile companies. The author analyzes data collection, data preparation, and feature selection using bivariate statistics on real mobile data. The process of training, testing and deploying decision-tree in the classification is detailed and illustrated through the case study. Our experiments show that our approach is promising for accurate analysis, identification and prediction of fraudulent customers. This research is helpful for intelligent and scientific management and decision making in fraud-related customers and activities, which benefits less economic losses and more efficient operations in the daily business life.
|