Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 53906 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Bagheri, Saeed; Namvar, Mehrzad
- Abstract:
- One of the important tasks of the customer relationship management of any company is to control the customer churn rate. Because according to research, retaining existing customers for a company costs less than attracting new customers. One of the most important tasks in order to control customer churn is to anticipate customers who intend to leave the company. This thesis proposes new method for customer churn prediction in mobile telecommunication industry. Mobile operators are structured in such a way that they can record and store a lot of data. In this dissertation, the most appropriate operator data was selected using data mining methods and various features related to customer churn are extracted from this data. Three categories of individual, trend_based and network_based features were extracted. Due to the small number of churners compared to loyal customers, we were faced with unbalanced data and to solve this problem, two approaches were used: Data balancing and Cost Sensitive Learning. Then, using the two mentioned approaches, Logistic Regression, random forest and artificial neural network were trained and evaluated their results. In the obtained results, the data balancing approach performed better than the Cost Sensitive Learning approach, as well as the accuracy of random forest and neural network algorithms was better than the Logistic Regression algorithm. Finally, best model predict Churners with top decile lift equal to 8.8232 and F1_Score equal to 0.6302
- Keywords:
- Customer Relationship Management ; Data Mining ; Churn Management ; Customer Churn Prediction ; Imbalanced Binary Classification
Digital Object List-
محتواي کتاب - view
Bookmark