Recommendation Feature of Scientific Articles on Open Journal System Using Content-based Filtering

OJS (Open Journal System) is an open-source platform for online journals, used by many people because it is easy to implement. OJS has various features, such as editor management, online submission, comprehensive indexing, etc. However, OJS does not have any recommendation feature to suggest suitable articles for users. This research aims to develop and embed a recommendation feature into the existing OJS to suggest relevant articles for users using Content-based Filtering approach, which focuses on the similarity of the data contents. Some parts of the article, such as the title, the keyword, and the journal scope are used as the reference data for the recommendation. K-Means clustering and Cosine Similarity are used as matching method, and the results of the recommendations are compared to find the most appropriate method to use. The result shows that the average precision score for the recommendations given by K-Means clustering method is better than that of Cosine Similarity method. The average precision score for K-Means clustering method is 68%, and the percentage of the recall is 64%. As for Cosine Similarity method, the average precision score is 44.15%, and the recall is 64%.