Treffer: Recommendation of Indian credit cards using topic modelling on rewards and offers.

Credit Card is one of the major cashless/digital payment options in India. Just by using the right credit cards for the right type of transactions, many users can save a lot of money and earn some extra money. The regular usage of credit cards and timely payments of bills by the users also help in improving/maintaining good credit scores. Credit card issuers provide a lot of offers and benefits based on their usage with certain kinds of merchants, modes of payment, vendors etc., The offer details are mentioned on the respective bank websites for each credit card. It's not feasible for customers to go through all the offers of all credit cards and pick the best ones which suit their criteria. The credit card offer details are collected from all the major bank websites using Selenium web driver and beautiful soup python library. The rewards and offers data are processed and analyzed using many Text Analytics methods like Exploratory Data Analysis (EDA), Word Frequency Distribution and Word Cloud. In this study, different topic modelling techniques are used and based on the performance best technique is chosen. The two conventional approaches Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) and the two latest novel approaches like BERTopic and Top2Vec are used. Their pros and cons and the results are evaluated. Based on the final analysis a credit card recommendation system is built. This system helps the customers to pick the best-suited credit cards as per their interests. The solution is deployed as a chatbot. [ABSTRACT FROM AUTHOR]