A machine learning based model for customer churn prediction in telecommunication

Customer churn is a major problem and one of the biggest problems for large firms. Companies are developing techniques to predict probable customer churn since it directly affects their revenue. In order to decrease customer turnover, it is crucial to identify the variables that contribute to this churn. This paper’s key contribution is to showcase the importance of customer churn in telecom that helps telecom providers identify consumers that are most likely to experience churn. The work is described in this study applies machine learning methods on datasets to predict whether a customer is likely to churn or not. To evaluate the effectiveness of churn prediction models, researchers have focused on assessing the accuracy of various machine learning models. Another significant contribution is the use of hyper-parameter tuning to increase the efficiency of the best resulted model. The accuracy of the model was improved by hyper-parameter tuning from 80.17% to 80.31%. The model is constructed and evaluated on the python platform using a sizable dataset that is produced by converting massive raw data provided by the Telco, a fictional telecommunications business. The model tested five different machine learning algorithms: Logistic Regression, Naive-Bayes Classifier, Support Vector Classifier (SVC), Decision-tree Classifier and Random Forest Classifier. However, using the Logistic Regression method produced the greatest results.