tcc2vec: RFM-informed representation learning on call graphs for churn prediction

Open Webpage

Sandra Mitrovic, Bart Baesens, Wilfried Lemahieu, Jochen De Weerdt

Published: 2021, Last Modified: 07 Oct 2025Inf. Sci. 2021EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We propose tcc2vec: An integrated framework for churn prediction (CP) in Telco.•We develop node2vec-inspired learning algorithm, scalable on very large graphs.•We devise novel RFM-informed call graph architecture and include its dynamic aspect.•We perform an in-depth experimental evaluation based on real-life datasets.•We provide guidelines for an adept representation learning on call graphs for CP.