Development of an AI Domain in a Data Mesh Network for Customer Credit Classification Using Transaction Data

According to the data mesh scheme for providing financial services, a data product based on customer transactions from the last 90 days is created within the operational domain. To obtain the data product, operational data corresponding to the open banking data model are anonymized, grouped, categorized, and further synthesized using GAN (Generative Adversarial Networks). The created data product is then transferred to the AI domain to determine customer credit ratings. This work investigates the operation and functionality of the AI domain, where the TensorFlow framework is used to create a neural network model. It is shown that the neural network model does not achieve maximum classification accuracy during training and testing but demonstrates high Area Under the Curve (AUC) of the ROC (Receiver Operating Characteristic) curve. This indicates that the model effectively separates classes while maintaining high sensitivity in correctly detecting both positive and negative samples. Integrating this model with traditional banking processes for customer credit classification within the data mesh structure will enable a more reliable and comprehensive credit management system.