Different Approaches to Recommender Systems

Hansda, Sadhana (2022) Different Approaches to Recommender Systems. Masters thesis, Indian Institute of Science Education and Research Kolkata.

Text (MS dissertation of Sadhana Hansda (17MS035)) 17MS035_Thesis_file.pdf - Submitted Version Restricted to Repository staff only Download (1MB)

Abstract

Recommendation systems are algorithms for suggesting products to customers. It helps customers find products they otherwise would have never searched for. Another way customers find a new product is through product bundling. Bundle recommender systems combine the concepts of recommender systems and product bundling to create an algorithm that recommends product bundles to customers. These are important m-commerce tactics that encourage customers to spend more money, resulting in increased profits for the company. This project mainly includes two approaches: collaborative filtering using Spectral Graph Clustering on MovieLens dataset and a Bundle recommender system with personalized pricing on a subset of “Steam m Video Game and Bundle Data”. The first model predicts product ratings based on prior user ratings and then they are compared with ratings predicted using cosine similarity and Pearson coefficient using RMSE score. Whereas the second model offers K product bundles using Jaccard similarity, similarity measure of collaborative filtering and personalized demand function.

Actions (login required)

View Item