<![CDATA[Public policy initiatives often trigger massive shifts in digital public opinion, such as the Free Nutritious Meal Program (MBG), which has garnered extensive attention from the Indonesian public on social media. Sentiment analysis serves as a vital instrument to map public opinion trends, particularly when dealing with large-scale, unstructured, and heterogeneous textual data. This study aims to analyze the distribution of public sentiment toward the MBG Program and evaluate the effectiveness of the lexicon-based method and Support Vector Machine (SVM) algorithm in classifying opinion texts. The dataset was collected from Twitter (X) via the Kaggle platform, comprising 10,524 public comments. The methodology begins with text preprocessing, including cleaning, case folding, tokenization, normalization, stopword removal, and stemming. Sentiment labeling was performed automatically using a lexicon-based approach referring to the InSet Lexicon to categorize data into three classes: positive, negative, and neutral. Subsequently, text representation was conducted using the Term Frequency–Inverse Document Frequency (TF–IDF) method and classified using an SVM model with a nested cross-validation scheme to maintain performance stability. The results indicate that public opinion is dominated by neutral sentiment at 48.1% (5,066 data points), followed by positive sentiment at 30.8%, and negative sentiment at 21.0%. This dominance of neutral sentiment reflects an informative, descriptive, and cautious public stance toward a policy still in its early implementation stages. Evaluation of the SVM model demonstrates highly stable and reliable performance, achieving an accuracy of 89.26%, with precision, recall, and F1-score each at 89%. This study concludes that the combination of lexicon-based automatic labeling and SVM is effective for public policy sentiment analysis, providing insights into public expectations and concerns regarding government programs.]]>
Copyrights © 2026
