A Privacy-Preserving Book Recommendation System for Electronic Libraries Using Federated Neural Collaborative Filtering
Abstract
The rapid increase in digital resources for electronic library systems has necessitated intelligent recommendation mechanisms that can help users discover relevant academic materials promptly and efficiently. Traditional recommendation systems are often based on centralized machine learning approaches that collect and store users' interaction data on a central server, thereby raising concerns regarding privacy, data security, and regulatory compliance. This study presents a Federated Learning-Based Book Prediction and Recommendation System for Electronic Library Environments. The proposed framework employs a privacy-preserving distributed learning architecture in which user reading histories, ratings, search activities, and borrowing patterns remain on local devices, but only model parameters are shared with a central aggregation server. A Neural Collaborative Filtering (NCF) model is adopted to learn user preferences and book characteristics, and the Federated Averaging (FedAvg) algorithm is utilized to aggregate locally trained models into a global recommendation model. An electronic library dataset comprising 5,000 users, 1,000 books, and over 200,000 user-book interactions was generated to simulate realistic library usage. The dataset was preprocessed and distributed among twenty federated clients for decentralized model training. The simulation was carried out using Python and PyTorch, while SQLite was employed for local data management, and Tkinter was used to develop a graphical user interface for the prototype application. Performance evaluation demonstrated that the proposed federated recommendation model achieved an accuracy of 92.84%, precision of 91.63%, recall of 90.78%, and F1-score of 91.20%. The study demonstrates that Federated Learning provides an effective framework for developing privacy-preserving recommendation systems capable of delivering accurate and personalized book recommendations in electronic library environments.
Keywords:
Federated Learning, Electronic Library System, Book Recommendation, Neural Collaborative Filtering, Privacy-Preserving Machine Learning, Federated AveragingDownloads
ACCESSES
Published
Issue
Section
License
Copyright (c) 2026 Oni Olujimi Olakunle, Osifade, Muyiwa Samuel, Aderanti Aderemi Taiwo, Adediran, Olatunji Sanmi (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.









