Recommendation System using NLP and Collaborative Filtering

Abstract

A significant amount of online shopping has been observed during the pandemic. Surveys suggest that 20-30% more businesses are moving online in response to the pandemic. The goal was to develop an apparel e-commerce website that is also tailored to the individual using Natural Language Processing and Collaborative Filtering algorithms, thereby recommending items based on what they choose. In the current system, the market uses a collaborative filtering and content-based filtering, hybrid model. We aimed to resolve the cold start problem that persists in many of the current e-commerce websites. The recommendation engine is based on the input given by the pre-trained Natural Language Processing model, where the items from the dataset are segregated into broad themes based on the product description. These themes are then put into clusters based on their similarity. The recommendation engine uses collaborative filtering and picks up similar products from the clusters formed to provide the users with appropriate recommendations.