An application of user segmentation and predictive modelling at a telecom company

Abstract

“The squeaky wheel gets the grease” is an American proverb used to convey the notion that only those who speak up tend to be heard. This was believed to be the case at the telecom company I interned at – they believed that while those who complain about an issue (in particular, an issue of no access to the service) get their problem resolved, there are others who have an issue but do not complain about it. The latter are likely to be dissatisfied customers, and must be identified. This report describes the approach taken to address this problem using machine learning. Unsupervised learning was used to segment the customer base into user profiles based on their viewing behaviour, to better understand their needs; and supervised learning was used to develop a predictive model to identify customers who have no access to the TV service, and to explore what factors (or combination of factors) are indicative of this issue.