Implementation of Content-Based Filtering in a Novel Recommendation System to Enhance User Experience

This study addresses a critical challenge in digital novel platforms: the difficulty of delivering personalized and accurate recommendations due to limited user interaction data. This limitation often leads to irrelevant or generic suggestions, which can diminish user engagement and hinder content discovery. The significance of solving this issue lies in enhancing user experience by ensuring that readers are presented with novels that truly align with their interests, even in the absence of extensive behavioral data. To overcome this problem, the study proposes an innovative hybrid recommendation system that integrates Content-Based Filtering (CBF) with the Random Forest algorithm. The system generates personalized recommendations by analyzing novel attributes such as title, genre, score, and popularity. The methodology involves extracting features from textual data using Term Frequency-Inverse Document Frequency (TF-IDF), followed by the calculation of cosine similarity to assess title relevance. These similarity scores are then combined with popularity predictions derived from the Random Forest model to produce final recommendations that reflect both content similarity and statistical relevance. The proposed system demonstrates strong performance, achieving an accuracy of 94.0%, precision of 81.4%, recall of 80.3%, and an F1-score of 80.8%. These results underscore the system’s capability to deliver accurate and diverse suggestions. By enhancing personalization and addressing the limitations of conventional CBF systems, this hybrid approach offers practical value for digital novel platforms. It serves as an effective tool for improving content discovery, increasing reader satisfaction, and supporting user retention in content-rich environments.