Sentiment Classification of Public Perception on LHKPN Using SVM and Naive Bayes

The public’s perception of the State Officials’ Wealth Report (LHKPN) serves as a vital measure of confidence in the government's commitment to transparency and efforts to combat corruption.This research seeks to examine public sentiment as reflected on the social media platform X. A dataset comprising 1,200 tweets was gathered and processed through various text mining methods, such as case folding, data cleaning, tokenization, normalization, stemming, stopword elimination, and TF-IDF vectorization. The tweets were then manually annotated into two sentiment categories: positive and negative, with 77.3% of tweets labeled as positive and 22.7% as negative. Sentiment classification was conducted using two machine learning algorithms: Support Vector Machine (SVM) and Naive Bayes. The Naive Bayes algorithm recorded an accuracy of 86.66%, with a precision of 0.93, a recall score of 0.88, and an F1-score of 0.87. Conversely, the SVM model with a linear kernel demonstrated superior performance, achieving an accuracy rate of 93.33%, along with a precision of 0.93, recall of 0.98, and an F1-score of 0.95. To uncover frequently occurring topics, WordCloud visualizations were generated. These revealed that positive tweets often included words such as ‘lapor’ and ‘transparan’, while negative ones were more likely to contain terms like ‘bohong’ and ‘korupsi’. These findings indicate that public sentiment toward the LHKPN initiative is largely favorable, despite persistent concerns surrounding integrity and trustworthiness in asset reporting. This study highlights the effectiveness of sentiment analysis in gauging public opinion and informing future policy improvements.