Predictive Modeling for Telco Customer Churn using Rough Set Theory

• Main Authors: Mokhairi, Makhtar, Mohd Nordin, Abdul Rahman, Mohd Khalid, Awang
• Format: Article
• Language: English; English
• Published: Asian Research Publishing Network 2016
• Subjects: HF Commerce; QA75 Electronic computers. Computer science
• Online Access: http://eprints.unisza.edu.my/7278/1/FH02-FIK-16-05877.pdf; http://eprints.unisza.edu.my/7278/2/FH02-FIK-16-05881.jpg; http://eprints.unisza.edu.my/7278/

A rough set is a mathematical tool to handle imprecise and imperfect information. It has been increasing in popularity recently in Knowledge Discovery in Database (KDD) and Machine Learning application. Rough set is one of the techniques used in KDD data mining. Data mining is an approach to extract useful information from a massive database for business purposes, for example, classifying customer churn. Churn is customer behaviour to terminate a service in favour of a competitor. Identifying customers who are likely to churn in the early stage will help firms to increase profitability since acquiring new customers is costly compared to retaining existing one. Limited research in investigating customer churn using machine learning techniques had led this research to discover the potential of rough set theory to enhance customer churn classification. This paper proposes a rough set predictive classification framework for customer churn in Telecommunication Companies. Experimental results show that the classification model is able to classify up to 83% to 98% accuracy for customer churn dataset. Overall, this indicates that the rough set theory is effective to classify customer churn compared to traditional statistical predictive approaches.