# Highlights
- This is a banking customers, accounts, products and interactions dataset that can be used for training and testing of GNN models to recommend products to customers
- The bank is assumed to be large retail bank and the retail product portfolio is considered for the data generation
- Synthetic data of 100K customers, 200K accounts, 42 products and 300K interactions (edges) can be generated. Effort has been taken to ensure the synthetic data is representative of the actual production data to the extent possible
- The code is parameterized to control the customer/account/product/interactions characteristics and volume
- Master data used – Master data like customer segment, city, customer risk, customer gender, product type etc is prepared manually with values taken to resemble a typical large retail bank. This data is used in generation of customers, accounts, products and interactions
- I have used this data in a Heterogenous Graph Transformer (HGT), a GNN model to…
