Telecom Customer Churn Prediction Model for Ethiopian Telecommunication

Abstract

Machine learning is a valuable tool for industries that deal with large amounts of data and want to enhance their operations. Ethio Telecom is one of these industries, which collects, processes, and stores huge volumes of records regularly. A major challenge for this industry is customer churn, which occurs when customers switch to another service provider and cause a loss of revenue. In industries that are based on subscriptions and face high competition and dynamism, customer churn is a serious problem. The telecommunication industry uses customer relationship management (CRM) databases to store a lot of data that contains hidden knowledge and patterns that can help them understand their customers better and make smarter decisions. However, because of the large amount of data that accumulates over time, CRM needs to apply business intelligence that can examine the causes and behaviors of churn customers from the existing data. The goal of this research is to create a machine-learning model that can classify customer data as churner or non-churner based on the customer's past data. The dataset for the experiment is obtained from Ethiopian telecommunication. The proposed model used random forest, support vector machine, logistic regression, and K-nearest neighbors’ algorithms toclassify customer data. A confusion matrix was used to measure the accuracy, precision, recall, and f1 score and assess the performance of the model. The research results showed that the RF classifier achieved an accuracy of 96.6%, SVM achieved an accuracy of 94.55%, the KNN classifier achieved an accuracy of 92.88% and LR achieved an accuracy of 87.25%. Based performance comparison of these classification algorithms, the best classifier for the customer churn classification problem is random forest.