Ensemble Book Recommendation System

Abstract

To address the information overload issue faced by users, recommendation systems have been introduced. However, when collaborative and content-based filtering approaches are used individually, recommendation systems face various issues such as cold start and inaccurate recommendations. This project aims to create an ensemble book recommendation system to mitigate these existing issues in recommendation systems. Hence, generating accurate and relevant book recommendations to improve user satisfaction. An ensemble book recommender system using both collaborative and content-based filtering approaches was developed using stacking ensemble method. LightGCN and SVD were the collaborative filtering approaches whereas TF-IDF was the content-based filtering approach utilized. To ensure data quality, data preprocessing was carried out. The individual models and the stacked model were evaluated to understand the performance of each model.