<![CDATA[This study aims to cluster student feedback on lecturers in the Academic Information System (SIAKAD) of Universitas PGRI Kanjuruhan Malang using the K-Means algorithm. We analyzed 33,229 comments through text preprocessing stages (normalization, tokenization, stopword removal, and stemming). Subsequently, we transformed the comments into numerical representations using the Term Frequency-Inverse Document Frequency (TF-IDF) method. We determined the optimal cluster number via the Elbow method, yielding three clusters. Evaluation of cluster quality using the Silhouette Score revealed a relatively low value (0.084). Nevertheless, further analysis with Latent Dirichlet Allocation (LDA) and lexicon-based sentiment analysis uncovered unique characteristics for each cluster: (1) dominated by general appreciation (good, friendly), (2) focused on learning aspects (clear material, teaching methods), and (3) containing personal comments and specific criticisms (excessive assignments, slow grading). These findings demonstrate that comment clustering complemented by topic modeling and sentiment analysis can provide strategic insights for the university to support academic decision-making (e.g., lecturer development programs) and enhance teaching quality in a targeted manner.]]>
Copyrights © 2026
