Churn Prediction using hierarchical generative models

Predicting customer churn creates opportunities to target customerswith a marketing action or a promotion to prevent them from leaving.In this paper, a hierarchical generative approach will be appliedin the context of churn prediction. The dependent variable and thecovariates are modeled jointly conditioned on a deep latent structure,which resembles the hidden structure in neural networks. The conditionallatent structure is capable of handling missing data and combiningheterogeneous data types. Latent structures with multiple layersare non-linear and can model complex dependencies between the independentvariables and the risk to churn. The hierarchical generativeapproach makes use of deep exponential families (DEFs). This classof models is able to extract a hierarchy of dependencies between latentvariables. Similar to deep unsupervised feature learning, this analysiscan improve predictions and provide extra insights into the natureof the data. The hidden layers in the DEFs enable the exploration ofinteresting structures in datasets. These patterns could help sales representativesin classifying customers according to their risk of churn,so that companies or managers can take more well-informed decisions.Predicting customer churn creates opportunities to target customerswith a marketing action or a promotion to prevent them from leaving.In this paper, a hierarchical generative approach will be appliedin the context of churn prediction. The dependent variable and thecovariates are modeled jointly conditioned on a deep latent structure,which resembles the hidden structure in neural networks. The conditionallatent structure is capable of handling missing data and combiningheterogeneous data types. Latent structures with multiple layersare non-linear and can model complex dependencies between the independentvariables and the risk to churn. The hierarchical generativeapproach makes use of deep exponential families (DEFs). This classof models is able to extract a hierarchy of dependencies between latentvariables. Similar to deep unsupervised feature learning, this analysiscan improve predictions and provide extra insights into the natureof the data. The hidden layers in the DEFs enable the exploration ofinteresting structures in datasets. These patterns could help sales representativesin classifying customers according to their risk of churn,so that companies or managers can take more well-informed decisions.C