OPTIMIZED PERFORMANCE FOR CONSUMER CHURN PREDICTION USING DATA-MINING FRAMEWORK

Abstract

Predicting whether a company will churn or not, is one of the most important tasks to be done in a business competitive environment, regarding keeping the consumers. We present an optimized data-mining framework to predict and proactively address customer churn in the consumer industry. We deploy a combination of machine learning techniques to predict churn pattern. Additionally, we present a novel feature selection measure that helps us to predict churn and enables better computational efficiency and reduced prediction time. In addition, to combine the predictions of various models, we employ ensemble learning. Experimental results reveal that our framework significantly outperforms traditional methods with 89.03% accuracy, 85.78% Precision, 91.14% Recall, and 95.07% f1-score on a real-world consumer churn dataset. We are proposing a framework where businesses can proactively catch customers at risk and go for prevent retention measures which can enhance customer retention and revenue as well.