A Fair-Generative approach for Customer Relationship Management

The advent of generative models has emerged as a fundamental shift of perspective in modern machine learning, with the potential to revolutionize a variety of domains. The generation of data that closely resembles real-world events has enabled the transformation of various businesses, increasing the possibility of building newer and more successful ones. In this work, we demonstrate the design and implementation of a machine-learning pipeline for financial CRM (Customer Relationship Management), specifically for predicting credit customer churning. The developed end-to-end pipeline proposes a predictive model with a high predictive capacity of recognizing churning behaviours that adheres to transparency and discriminant-free paradigms. The approach combines a state-of-the-art tabular generative procedure exploiting Generative Adversarial Network architecture to mitigate the dataset's lack of representative samples with major company service solutions such as IBM DataStage and H2O AutoML for effectively assessing the data ingestion and model creation phases.