Research On The Application Of Date Mining For Telecom CRM

Facing the fierce competition in telecommunication market and development of information technology, telecom companies must establish a management pattern with"centered on customers". Therefore, it's a significant aspect for telecom company to improve CRM level using data mining technology to mine and analyze a great deal of telecom companies'customer data and discover various of potential, valuable and regularity knowledge. It has theoretical meaning and applied value. Aimed at some application problems of data mining in telecom CRM, the thesis takes the study with the methods of theoretical analysis and empirical research. The object matter is as follows:1. This thesis analyzes the actuality of telecom companies'IT system in detail and establishes a telecom data warehouse. Then, the analysis subject, data models (physical model and logical model) and the implementation method of data warehouse is researched.2. The thesis studies the CRM theory by the numbers; designes the closed-loop and four-layer architecture of telecom CRM. Telecom customer management is researched by the numbers, and a telecom customer lifecycle management model based on data mining is brought forward under the phase of telecom customer lifecycle management theory.3. The three dimensions telecom customer value model based on current value, incremental value and perserved value is brought forward according to CLV theory. Based on this theory, telecom customer value evaluation index system and computing method combined with AHP method are put forward, and empirical research on the telecom company PHS customers is taken.4. A modified K-means clustering model optimized by genetic algorithms is established. In the data process and in results evaluation this thesis researches supervised clustering model evaluation method. Finally, this thesis takes an empirical analysis with the data of public customers in a telecom company and describes the character of clusters of customers.5. According to the cost-sensitive learning theory, this thesis makes use of under-sampling and AdaCost to build telecom customer churn prediction model with minimum-cost, and demonstrates that minimum-cost prediction model is better than traditional model through total cost and model benefit analysis.The thesis provides beneficial reference to telecom companies in analyzing customer behavior and advancing CRM level using data mining technology, and shows important meaning in theoretical research and engineering practice.