A competing risks model based on latent Dirichlet Allocation for predicting churn reasons

Due to low switching costs and stiff competition, customer relationship management has become a central component in the marketing strategy of telecommunication service providers. Since acquiring new are five times higher than maintaining an existing customer, providers eager reduce churn rate. A solid understanding behavior can help address this problem. Reducing rate translate into significant revenue gains might provide edge outperform competitor. In paper, we predict propensity for customers Dutch provider by employing duration model. While predicting churn, simultaneously reason which churns, using competing risks valuable textual data based on transcripts calls between center, incorporate topics extracted from as variables our models, Latent Dirichlet Allocation (LDA). We compare four models find that have incorporated topic usually yield best forecasts. Also, investigated beat considered benchmark model, is model currently deployed at provider.