Developing a Content-Based Book Recommendation System for School E-Libraries Using TF-IDF and Cosine Similarity

Abstrak

The rapid growth of e-libraries has increased the availability of digital book collections, making it difficult for users to identify relevant reading materials. This study develops an automated book recommendation system for the SMKN 1 Semen e-library using Content-Based Filtering with TF-IDF for term weighting and Cosine Similarity for similarity measurement. The dataset consisted of 35 book records containing metadata and book descriptions collected from the SMKN 1 Semen e-library. Text data were processed through case folding, tokenization, stopword removal, and stemming before recommendation generation. The system was developed using the Laravel framework and a MySQL database following the Research and Development (R&D) method with the Waterfall model. Recommendation performance was evaluated using a Top-5 recommendation scenario, where the recommended books were compared with manually identified relevant books based on content similarity using Precision and Recall metrics, while system functionality was verified through Black Box Testing. The experimental results achieved a Precision of 80% and a Recall of 57.14%, indicating that the proposed approach effectively generates relevant book recommendations. This study contributes by demonstrating that the integration of Content-Based Filtering, TF-IDF, and Cosine Similarity provides an effective recommendation approach for school e-libraries with limited user interaction data, enabling personalized book recommendations and improving the efficiency of book discovery in small-scale digital library environments.