Personalized Book Intelligent Recommendation System

Abstract - As libraries undergo digital transformation, accompanied by advancements in information technology in academic libraries, it is evident that readers will want different and personalized services in addition to checking out books. Along with meeting the changing expectations of users, this research proposes an enhanced item-based collaborative-filtering recommendation algorithm that employs an average model representation to improve the accuracy measurements, as well as the use of Neural Collaborative Filtering (NCF) for modeling the user-item interaction by utilizing deep neural networks to improve the expressiveness of recommendations in a non-linear way using NCF instead of linear factors. The enhanced library recommendation system is developed using Django (v4.1.2), a high-level web framework that creates an organized way to develop a backend application. The library recommendation system uses Pandas (v1.5.0) and NumPy (v1.23.3) to process the behaviour data of users and perform an analysis of how users behave in the library system. The requests (v2.28.1) and requests-oauthlib (v1.3.1) packages were used to interact with the APIs. Gunicorn (v20.0.1) is used as the WSGI server for the applications production environment. Environment management was done with virtual (v20.13.0) in the backend and nodeenv (v1.6.0) in the frontend. The ability to personalize the recommendations of the library service to the user in the proposed approach is confirmed with the significant improvement in accuracy.[7][8] Keywords: Collaborative Filtering; Neural Collaborative Filtering; Recommender System; University Library; Personalized Services; Django; Machine Learning; Data Analysis; Web Application