<![CDATA[The assignment of thesis supervisors is a critical academic decision that directly affects research quality and completion outcomes. However, supervisor selection in many higher education institutions remains reliant on subjective judgment and manual inspection of lecturers’ research profiles. This study proposes a content-based thesis supervisor recommendation system that integrates research interest clustering and cosine similarity to support more objective and transparent supervisor assignment. Lecturers’ research interests are derived from publication titles and abstracts collected from Google Scholar and represented using TF–IDF weighting. K-means clustering is applied to model dominant research interest themes, while cosine similarity is used to match students’ thesis proposal texts with clustered publication data. The proposed approach was implemented as a web-based decision-support system and evaluated using publication data from 21 lecturers comprising 469 records. The results indicate that research interest clustering provides a structured and interpretable representation of academic expertise, enabling contextually relevant supervisor recommendations. The system demonstrates practical value by enhancing transparency, consistency, and efficiency in academic decision-making. This study contributes to applied research on academic recommendation systems by extending publication-based approaches through cluster-level modeling of research interests.]]>
