PENERAPAN CONTENT-BASED FILTERING PADA SISTEM REKOMENDASI NOVEL UNTUK PENINGKATAN PENGALAMAN PENGGUNA

This research aims to develop a novel recommendation system capable of delivering relevant and personalized reading suggestions to users based on their preferences.
The method employed is Content-Based Filtering, which analyzes novel content attributes such as title, genre, and author. To improve the accuracy and quality of recommendations, the Random Forest algorithm is integrated to predict a novel’s popularity based on its rating and other attributes. The dataset used in this study consists of 4,500 publicly available novel titles, with features such as genre, score, popularity, and release year. The research process involves data cleaning, feature extraction using TF-IDF, text similarity computation via cosine similarity, and model training using Random Forest classification. Model evaluation is conducted using a Confusion Matrix, achieving 94,2% accuracy, 78,3% precision, 87.4% recall, and an F1-score of 82.6%. These results demonstrate that the integration of
Content-Based Filtering and Random Forest can produce more accurate and user-aligned recommendations. The system is then implemented as a web-based application using Streamlit, allowing users to interactively access the
recommendation features. The application includes three main features: recommendations based on popularity, rating score, and similarity in genre and title. This approach is expected to enhance user reading experiences by offering more relevant, precise, and satisfying novel suggestions, especially in the digital platform ecosystem.
Keywords: Recommendation System, Content-Based Filtering, Random Forest, TF-IDF, Cosine Similarity, Streamlit, Novel.