<![CDATA[Content-based filtering movie recommendation systems are widely used to help users find relevant movies. However, most studies still rely on a single feature such as a synopsis, resulting in a less informative feature representation that impacts the quality of recommendations. This study aims to improve the accuracy of movie recommendation systems using TF-IDF and Cosine Similarity through the application of Hybrid Feature Engineering, which combines synopsis, genre, and keywords. Evaluation was conducted using the TMDb 5000 Movie Dataset with Precision, Recall, and F1-Score metrics in a Top-10 Recommendation scenario. The results showed that the hybrid model increased Precision from 0.7600 to 0.9600 (26.3%), Recall from 0.0048 to 0.0060 (25.0%), and F1-Score from 0.0095 to 0.0120 (26.3%) compared to the baseline model. The system was also successfully implemented as a Streamlit-based web application. These results indicate that combining multiple features through Hybrid Feature Engineering can produce a more informative feature representation and improve the quality of recommendations in content-based filtering systems]]>
Copyrights © 2026
