CUSTOMER CHURN PREDICTION IN TELECOM USING MACHINE LEARNING AND DATA MINING

Abstract

Customer churn prediction is crucial in reducing the loss of customers and enhancing the retention strategies by telecom companies. This paper suggests a machine learning-based system in the case of the IBM Telco Customer Churn data to identify the probability of a customer switching the service. The methods of data preprocessing, including dealing with missing values, coding of categorical variables, log transformation, and feature scaling are used to improve the quality of data. Exploratory Data Analysis (EDA) will be performed to find some trends and factors that are important in churn. There are several supervised learning models, such as Decision Tree, Random Forest, and X GBoost that are implemented and evaluated. Random Oversampling is used to deal with the issue of class imbalance in order to enhance the model performance on minority class examples. Training and testing accuracy is used to evaluate the models with the ensemble models (Random Forest and XG Boost) performing well and generalizing better than the Decision Tree model. The findings show that the type of contract, technical support and payment method are important factors influencing the customer churn, which means that machine learning techniques are quite helpful in the customer retention strategies of the telecom industry.