Feature Selections and Classification Model for Customer Churn

• Main Authors: Mokhairi, Makhtar, Mohd Khalid, Awang, Mohd Nordin, Abdul Rahman
• Format: Article
• Language: English; English
• Published: Asian Research Publishing Network 2015
• Subjects: HB Economic Theory; QA75 Electronic computers. Computer science
• Online Access: http://eprints.unisza.edu.my/6027/1/FH02-FIK-15-03308.pdf; http://eprints.unisza.edu.my/6027/2/FH02-FIK-15-03462.jpg; http://eprints.unisza.edu.my/6027/

As customers actively exercise their right to change to a better service and since engaging new customers is more costly compared to retaining loyal customers, customer churn has become the main focus for one organization. This phenomenon affects many industries such as telecommunication companies which need to provide excellent service in order to win over the competition. Several models were developed in previous research using various methods such as the conventional statistical method, decision tree based model and neural network based approach in predicting customer churn. Several experiments were conducted in this research for feature selection and classification from selected customer churn dataset to compare its usefulness among the different feature selections and classifications using a data mining tool. The results from the experiments showed that the Logistic Model Tree (LMT) method is the best method for this dataset with a 95% accuracy enhanced using neural network from previous research.