Autonomous toolkit to forecast customer churn

Customer churn prediction is one of the most important requirements in customer relationship management. It aims to retain valuable customers in order to maximize the profit of a company. In this paper, we propose an autonomous toolkit to forecast customers churn (ATFC) — an autonomous customer churn toolkit which predicts churning behavior of customers in the telecom industry. ATFC gives a customer churn prediction model which can fit generally in similar kinds of problems and datasets. It predicts which customers are at the risk of leaving the company. It is important for managers of the telecom industry to retain their loyal customers for the growth of their company and for improving their customer relationship management factor. ATFC accomplishes this task with the help of the most popular machine learning algorithms which were applied to the challenging problem of the customer churn in the telecom industry. We have used telecom company based dataset of BigML repository. Therefore, popular machine learning algorithms such as Decision Trees, Logistic Regression, Random Forest and Gradient Boosting were used to develop a model that can predict telecom customer churn efficiently and effectively.The results revealed that Random Forest outperforms by exhibiting low churn rate. The analysis of the algorithm was carried out based on ROC curve Precision-Recall and F-measure.The churn dataset analysis revealed that there are 10 features which causecustomer to churn. This prediction informs companies which features and services they need to target and improve.