Classification of Telecommunication Customers based on Profitability: A Supervised Machine Learning Approach

Abstract

In this modern world the telecommunication industry plays an important role. At the same time, the telecommunication industry is facing a serious problem of losing its potential customers. Customer relationship management is the combination of practices, strategies and technologies that companies employ to manage and analyze customer interactions and data throughout the customer lifecycle. Customer retention is one of the most difficult challenges in the telecommunications industry. Most researchers focus on developing suitable models to classify customers based on their profitability levels for many different businesses. This study has proposed a supervised machine learning approach for the classification of telecommunication customers based on profitability. Three classes namely high profitable customers, low profitable customers, and average profitable customers were considered based on their purchasing behaviours. Around 10,000 post-paid subscriber details with 12 attributes were collected and analyzed in the study. As a result, the classification technique has been used to reduce the size of the feature set and to classify them based on profitability. Various supervised machine learning algorithms were applied on the data set to choose the best algorithm for developing the final prediction model. Decision Tree, LightGBM, Random Forest, Support Vector Machine, XGBoost, AdaBoost, Na¨ıve Bayes, CatBoost, Artificial Neural Network, and K-Nearest Neighbor algorithms were considered for the analysis. Out of 10,000 pre-paid subscribers, 5000 are high profitable customers, 3000 are low profitable customers, and the remaining 2000 are average profitable customers, with the Support Vector Machine algorithm outperforming all other classification algorithms. It generated the best results, yielding the highest accuracy (77.80%) while producing a low error rate of 22.20%. Accordingly, the Support Vector Machine algorithm was identified for developing the final prediction model for the telecommunication customer classification based on profitability.