Procedia Computer Science · 2021 · 33 citations · 30 references
ComputingBusiness AnalyticsData ScienceData MiningMachine LearningPredictive AnalyticsChurn AnalysisChurn ManagementCustomer Churn PredictionCustomer LoyaltyDeep LearningRepresentation LearningChurn DetectionChurn ModelingCustomer Churn ModelingExplainable Churn ModelingExplainable Churn PredictionPredictive PerformancePredictive Performance EvaluationVector Embedding
To face the tight competition in the telecommunication industry, it is important to minimize the rate of customers stopping their service subscription, which is known as customer churn. For that goal, an explainable predictive customer churn model is an essential tool to be owned by a telecommunication provider. In this paper, we developed the explainable model by utilizing the concept of vector embedding in Deep Learning. We show that the model can reveal churning customers that can potentially be converted back to use the previous telecommunication service. The generated vectors are also highly discriminative between the churning and loyal customers, which enable the developed models to be highly predictive for determining whether a customer would cease his/her service subscription or not. The best model in our experiment achieved a predictive performance of 81.16%, measured by the F1 Score. Further analysis on the clusters similarity and t-SNE plot also confirmed that the generated vectors are discriminative for churn prediction.
30
Laurens van der Maaten, Geoffrey E. Hinton · Journal of Machine Learning Research · 2008 · 35.7K citations
Distributed Representations of Words and Phrases and their Compositionality
Tomáš Mikolov, Ilya Sutskever, Kai Chen et al. · arXiv (Cornell University) · 2013 · 18.1K citations · Full text