Teoh Jay Shen, Abdul Samad Bin Shibghatullah
Institute of Computer Science & Digital Innovation, UCSI University, Cheras, Kuala Lumpur, Malaysia
pp. 85–91
ABSTRACT
Customer churn is a constant issue that poses a serious threat and is one of the top issues for telecom businesses. The businesses are working to develop and build a strategy to anticipate customer attrition. This is why it is important to identify the sources of client churn. Churn prediction is the process of identifying which consumers are most likely to stop using a service or to cancel their subscription. Because getting new customers frequently costs more than keeping existing ones, it is an important prediction for many firms. The suggested models built in this work use both deep learning and machine learning algorithms. These models were developed and tested using the Python environment and a publicly available dataset from www.kaggle.com. This dataset, which was used in the construction of the models' training and testing phases, includes 7043 rows of customer data with 21 features. Four different machine learning and deep learning algorithms were utilized by these models, including the Artificial Neural Network, Self-Organizing Map, Decision Tree and a hybrid model with the combination of the Self-Organizing Map and Artificial Neural network algorithms

ARTICLE INFO
Article History
Received 23 November 2021
Accepted 28 October 2022

Keywords
Machine learning
Deep learning
Churn prediction
Telco industry

js3206

Download article(PDF)