Finding relevant news is becoming increasingly
difficult for users. Currently, people often experience information
overload due to the abundance of online news sources available,
making it difficult for them to select news that interests them.
To address this issue, a news recommendation system is needed
to help users select news that aligns with their preferences.
Based on previous research, the developed recommendation
systems still have limitations, such as reliance on ratings or
features without dimensionality reduction, text vector similarity
measurement, or recommendations based solely on users’
previous behavior. Therefore, we propose a more comprehensive
approach by combining Collaborative Filtering, which builds
a model based on user behavior, and Content-Based Filtering,
which builds a model based on item descriptions and preference
profiles. We apply the Singular Value Decomposition method
to find the latent representation of user interactions, as well
as cosine similarity to measure the similarity between items
or users. The evaluation results show that the hybrid method
achieved a recall of 0.66, a precision of 0.40, and an F1-score of
0.46. Meanwhile, single methods such as Collaborative Filtering
achieved a recall of 0.53, precision of 0.32, and an F1-score
of 0.40, while Content-Based Filtering achieved a recall of
0.60, precision of 0.36, and an F1-score of 0.45. The results of
this study show that the hybrid method is effective in news
recommendation systems with good ability to identify relevant
information.