The loss of customers is an issue for businesses of all kinds. When a client leaves, it is a big deal for the business. The phrase "churn prediction (CP)" is used to describe the process of figuring out which customers are most likely to terminate a service subscription. For many businesses, this forecast is especially important because acquiring new clients can be more costly than maintaining current ones. There is a current problem with customer churn (CC) research and prediction in the telecommunications industry since it is crucial for these businesses to know which consumers are likely to cancel their service. Machine learning (ML) strategies and models play a big role for businesses in the present financial environment. This is because the cost of obtaining a new customer is higher than the expense of maintaining an existing one. This project focuses on several ML algorithms for forecasting customer turnover using the telecom CC dataset from Kaggle. Our proposed approach has six facets. The first two sides are dedicated to preliminary data processing and exploratory data analysis (EDA). After considering data balance with SMOTE in the third stage, the data is split into a trainset and a testset with a 70%:30% split. On the train set, we employed the most popular prediction models, including Extra tree, Decision Tree (DT), XGBoost, and Voting, and we compared their accuracy using ensemble methods. A final analysis of the test set results in python simulator was performed using accuracy, precision, recall, & the f1-score curve. It was found that Proposed Voting had the highest accuracy, at 85.63 percent. Finally, the results of the experiments illustrate how the suggested system performs in comparison to other systems.

Customer Churn Prediction, Telecom Industry, Machine Learning, Extra Tree, Decision Tree, Xgboost, And Voting Classifiers.