<![CDATA[The selection of a thesis supervisor is recommended based on the alignment between the student's research topic and the supervisor's academic expertise. Currently, the supervisor selection process often relies on manual methods based on subjective preferences or supervisor availability, which risks creating a mismatch between the supervisor's competencies and the research topic. This study aims to implement machine learning model Term Frequency–Inverse Document Frequency (TF-IDF) and Cosine similarity approach for recommending thesis supervisors to enhance the objectivity and efficiency of the academic process. The methodology involved collecting a dataset of research titles and abstracts from lecturers in the Health Administration study program at STIKes Budi Mulia Sriwijaya. This was followed by text preprocessing stages, including case folding, tokenization, stopword removal, and stemming. Subsequently, term weighting was calculated using the TF-IDF algorithm, and the semantic similarity between documents was measured using Cosine Similarity. The analysis results indicate a highest TF-IDF score of 95.81, signifying a high degree of topic focus. Meanwhile, the highest Cosine similarity score was 0.77. The visualization of results in a heatmap illustrates the clustering of relationships between titles based on their level of topical similarity]]>
Copyrights © 2026
