Explaining and predicting customer churn by monotonic rules induced from ordinal data

In the course of a computational experiment on bank customer churn data, we demonstrate the explanatory and predictive capacity of monotonic decision rules. The data exhibit a partially ordinal character, as certain attribute value sets describing the clients are ordered and demonstrate a monotonic relationship with churn or non-churn outcomes. The data are structured by the Variable Consistency Dominance-based Rough Set Approach (VC-DRSA) prior to the induction of monotonic decision rules. The supervised learning is conducted using an extended version of VC-DRSA, implemented in RuLeStudio and RuleVisualization programs. The first one is designed to experiment with parameterized rule models, and the second one is used for visualization and a thorough examination of the rule model. The monotonic decision rules give insight into the bank data, characterizing loyal customers and the ones who left the bank. Such an approach is in line with explainable AI, aiming to obtain a transparent decision model, that can be easily understood by decision-makers. We also compare the predictive performance of monotonic rules with some well-known machine learning models.