AI driven best credit card recommendation model Qorus Banking Innovation Awards 2022

Objective was to recommend the best/second/third best credit card using advance Machine learning technique out of gamut of cards eligible instead of showing top three cards which is rule based. In order to achieve the above, a 360-degree view of customers was created basis 100 Mn+ transactions across Debit Card, UPI, Net banking, Digital footprints, campaign response, applying advanced NLP techniques on the unstructured data such as Chats from Bots and Whatsapp conversations and CRM interaction. These helped to create personas of the customers whether customer is a Traveler, Foodie, Movie Buff, Shopaholic etc. Additionally, 20+ surrogates developed to overlay recommendations for selected credit cards basis dominant behavior and finally arrived best/second/third best card for customers using multi-class classification model. ● Methods utilized A two-stage analytical based solution was designed Stage 1 – Recommending top three credit cards using multi-class classification model Stage 2 – Identifying Savings on usage of card by calculating spend potential at customer level n=1 Tools and platforms • An in-house unified text classification engine using NLP algorithm was developed using Python to extract merchant category from 50Mn+ UPI & Net banking transactions • Used Multi-class classification ML model using Python to identify best/second/third best credit card from gamut of cards eligible • Developed Clustering model using Python to identify true spend potential at customer level n=1