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: 54039 (01)
- University: Sharif University of Technology
- Department: Industrial Engineering
- Advisor(s): Rafiee, Majid
- Abstract:
- Markets are becoming more saturated every day and competition between different businesses is increasing. The importance of managing Customer churn in various businesses has become increasingly important because the cost of attracting a new customer is many times greater than retaining an existing customer. With the development of data mining and its increasing expansion and the other side, the increase of stored information related to various organizations and businesses has accelerated the operations of extracting knowledge from data. Today, businesses are moving towards the use of intelligent knowledge extraction systems, of which Customer churn prediction systems are one of the most important. Using systems that record customer interactions, these systems extract patterns for predicting customer churn behavior and help business managers determine business policies to help retain customers.In this research, a method based on the method of aggregation of experts is presented in which there are three subsystems that determine the final output according to their output. In the proposed method, the three machine learning methods of decision tree, support vector machine and evolutionary neural network predict the input training data in parallel and the output of these three subsystems for each sample along with the input data of each sample enters a neural network to produce the final prediction. The neural network that produces the final response acts as an aggregation unit and, in addition to the diagnosis it makes using input sample information, uses the output of three other subsystems to generate the final response. Also, to improve the neural network subsystem, the Harris hawks optimization (HHO) has been used to perform the neural network training process (determining the weights of the neural network).The proposed method has been trained and tested using the Telco Customer Churn dataset and the comparisons performed showed that the proposed method showed high performance with 98.25% accuracy
- Keywords:
- Machine Learning ; Neural Network ; Decision Making Tree ; Support Vector Machine (SVM) ; Customer Loyalty ; Data Mining ; Harris Hawks Optimization (HHO)
Digital Object List
-
محتواي کتاب
- view
Bookmark