<![CDATA[The rapid advancement of digital technology has increased the need for intelligent systems to filter job vacancies that match user profiles. This study aims to develop a job recommendation system based on a combination of Cosine Similarity, SMART, and EDAS methods. Job data were obtained from the JobStreet website and processed through text preprocessing stages such as tokenization, stopword removal, and stemming. Job descriptions and job seeker profiles were converted into numerical vectors using the TF-IDF method. Cosine Similarity was used to measure content similarity, SMART to evaluate suitability based on weighted criteria such as education and experience, and EDAS to assess alternatives relative to the average solution. System evaluation was conducted using precision, recall, F1-score, and mean Average Precision (mAP) metrics. Results show that Cosine Similarity alone had the lowest performance (F1-score 41.9%, mAP 42.3%), improved with the addition of SMART (F1-score 51.1%, mAP 50.9%), and achieved the best results with the integration of Cosine Similarity and EDAS (F1-score 66.5%, mAP 65.8%). Therefore, the integration of text similarity and multi-criteria decision-making methods effectively enhances the accuracy and relevance of job vacancy recommendations.]]>
Copyrights © 2025
