Customer churn prediction in telecommunication industry through machine learning based Fine-tuned XGBoost algorithm

Abstract

Telecom companies require the accurate prediction of probable churn customers to improvise the customer relationship management; this is addressed through customer churn prediction model. Customer churn is one of the major issue faced by telecommunication companies due to various competitors and this further cause’s loss of revenue. Moreover, the customer churn prediction model helps in predicting the potential churners; hence plenty of research has been carried out in past and few of them succeeded through machine learning approach. However data imbalance remains the major issue, hence in this research work we design and develop Fine-tuned XGBoost model which addresses the issue of imbalance dataset through introducing the feature function; further, it also tackles with data sparsity and overfitting problem. Finetuned XGBoost is evaluated through comparing the machine-learning algorithm in terms of performance metrics like precision, recall and accuracy and fine-tuned XGBoost outperforms the existing model.