Personalizing the Feed: A Content-Based Recommendation System Framework for Indonesian News Portals
Year: 2026
Type: article
Abstract: This study aimed to address the difficulty of finding relevant news amid information overload on online news portals. The proposed solution was the development of a news recommendation system using the Content-based Filtering method, implemented in Indonesian. The system utilized the Term-Frequency-Inverse Document Frequency (TF-IDF) model for content weighting and Cosine Similarity to measure the similarity between news articles and user preferences. The dataset consisted of news articles from popular Indonesian news portals, such as Detik, CNBC Indonesia, Tempo, Kumparan, and Okezone, totalling 32,735 articles. The methodology involved data collection, text processing, TF-IDF for word weighting, and cosine similarity for content similarity. Evaluation was conducted using metrics such as precision, recall, and fl-Score to assess system performance. Testing results showed the system achieved a precision of 1.00 and an Fl-Score of 0.96, indicating strong performance in recommending relevant news based on user preferences. In conclusion, this system has proven effective and efficient for helping users find news based on their interests. Future developments could include additional features, such as integration with other methods to improve recommendation accuracy.
Institution: Universitas Bhinneka Nusantara
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7204495248