<![CDATA[Public policy in the education sector requires continuous evaluation to ensure the effectiveness of its implementation, one example being the Gratispol Program organized by the Provincial Government of East Kalimantan. Student opinions on this program are widely expressed through Instagram comment sections. However, the unstructured nature of the data and the presence of informal language make manual analysis difficult. This study aims to analyze student sentiment based on 539 comments from 21 posts on the official Instagram accounts of 14 higher education institutions in East Kalimantan. The data were initially labeled using IndoBERT, manually validated, extracted using TF-IDF, balanced using SMOTE, and then classified using a linear-kernel SVM algorithm optimized through GridSearchCV. The results show that neutral sentiment dominates (50.5%), followed by negative (28.9%) and positive (20.6%) sentiment. Parameter optimization increased the model's accuracy from 62.96% to 67.90%. Negative comments were dominated by complaints about delays in fund disbursement, while positive comments contained appreciation for the program. These findings provide an empirical picture of student perceptions that can serve as input for the Provincial Government of East Kalimantan in improving the implementation of the Gratispol Program.]]>
Copyrights © 2026
