Customer Churn in Telecom Sector: Analyzing the Effectiveness of Machine Learning Techniques

The number of customers that stopped using a company’s product or service during a particular time is known as customer churn. Businesses may prevent churn by taking preventative measures when they can anticipate it before it occurs. Specifically in the telecommunication sector due to various providers, there is great competition. To compete in the market, telecom firms provide all basic services, easy access to the Internet, quality phone service, etc., to all mobile users and still it is a challenge to hold the clients. Therefore, it is an important task to understand the customer needs of all age groups. So by a proper prediction of customer churn, companies can reduce the rate of churn by immediately taking action regarding it. In this study, the authors show different exploratory data analysis (EDA) between different parameters which could affect the churn. Going further, the data has been divided for training which is 80% of the whole and the remaining 20% is kept as the test data. By comparing various machine learning (ML) models such as SVM, KNN, XGBoost, decision tree, and random forest, the best-performing model for the dataset is identified as the RF model. The best precision obtained is with RF, with an accuracy of 82%, and the least precise was from KNN with an accuracy of 76%. The study will provide a detailed view of the problem of customer churn and how it can be controlled.