International Journal of Science, Engineering and TechnologyAuthors: Harsh N Sorathiya, Manthan Shah, Dhruv Shah, Ovesh Khatri, Professor Rahul Moud
Abstract: This paper delineates the design and implementation of an integrated platform for personalized book recommendations. The system is architected upon a decoupled three-tier model, featuring a dynamic user interface built with React.js and a robust backend service developed in Python-Flask. Central to the platform is a hybrid recommendation engine that synergizes item-item collaborative filtering—which employs Cosine Similarity on a pre-calculated similarity matrix—with a content-based fallback mechanism. This dual-strategy approach is specifically engineered to overcome the prevalent challenges of data sparsity and the cold-start problem. To ensure persistent personalization, user data and interaction histories are stored in a cloud-hosted PostgreSQL database and managed via the SQLAlchemy Object-Relational Mapper (ORM). Security is enforced through a stateless JSON Web Token (JWT) authentication protocol, which also underpins the system's role-based access control for administrative functions. This research provides a practical blueprint for the development of scalable, real-world recommender systems by synthesizing established algorithms with contemporary software engineering methodologies.
