Article recommendation system using content-based filtering / Muhammad Hazrul Afiq Kamaruddin

The challenges in the education industry include the time-consuming process of manually searching for relevant research articles, which reduces productivity and negatively impacts academic performance. Furthermore, the growing issue of information overload and anxiety among students and researchers raises the risk of burnout and decreases overall academic performance. In this study, article recommendation system using content-based filtering was designed and developed to address the challenges. The algorithm used is able to generate relevant article recommendations based on the content of the article. The algorithm consists of two components which are Term Frequency - Inverse Document Frequency (TF-IDF) and Cosine Similarity. TF-IDF calculates the weightage of user query and each keyword in each article. Vectors containing weightage values for both user query and articles in dataset will be calculated using Cosine Similarity to obtain similarity value. Articles recommendation will be generated after sorting and filtering based on threshold value. The result was evaluated using confusion matrix and evaluation metrics such accuracy, precision, recall and F1 score. The article recommendation system is able to achieve up to 99% accuracy, 86% precision, 76% recall and F1 score of 0.8 where the threshold value is 0.1. Overall, the project is successful as it is able to generate relevant articles accurately.