MCS: Multiple classifier system to predict the churners in the telecom industry

Abstract

Multiple classifiers for prediction or classification has gained popularity in recent years. Ensemble Technique perform best predictions as compared to traditional classifiers. This has resulted in the experimentation with new ways of ensemble creation. This paper presents a multiple classifier system (MCS) that can outperform traditional classifiers. Experiments are performed on a benchmark Customer Churn Dataset (available on UCI repository) and a newly created dataset from a South Asian wireless telecom operator. MCS achieved accuracies of 97% and 86% on the UCI churn dataset and private dataset, respectively. MCS as compared to existing best approaches realized the best results on the private and public datasets.