Understanding telecom customer churn with machine learning: from prediction to causal inference
Retention campaigns are used by telecommunication companies to prevent customer churn, but their effectiveness depends on the availability of accurate prediction models. In this work, we discuss three aspects of a real churn prediction problem in collaboration with Orange : the design of an accurate prediction model, the application of data-driven causal inference and the impact of causally relevant variables.
Théo Verhelst got his Bachelor in Computer Science (Grande Distinction) from ULB in 2017. Then he took part to the Erasmus program at Southampton University during his first year of the Master. He obtained his Master in Computer science (Grande Distinction) at ULB in 2019. He is interested in topics such as machine learning, causal inference, information theory, and probability.