ANALISIS SENTIMEN MASYARAKAT TWITTER TERHADAP KEBIJAKAN EFISIENSI ANGGARAN KEMENTERIAN MENGGUNAKAN SVM

Abstract

Sentiment analysis of the ministry's budget efficiency policy is crucial to understanding public responses to government policies. This study employs the support vector machine (SVM) method to classify positive and negative sentiments from 1,418 tweets collected through crawling using Twitter API v2 between February 10 and 22. The text processing steps include case folding, cleaning, tokenizing, stopword removal, stemming, and weighting using the term frequency-inverse document frequency (TF-IDF) method. The analysis results indicate that negative sentiment dominates over positive sentiment, reflecting public criticism and dissatisfaction with the policy. The SVM model was evaluated using k-fold cross-validation with k values ranging from 2 to 10, achieving the best accuracy of 94.76% with 10-fold validation. Evaluation using the confusion matrix showed a precision of 92.85%, a recall of 91.32%, and an AUC of 0.972, indicating excellent model performance in sentiment classification. These findings suggest that the SVM model is effective in analyzing public sentiment toward government policies and can be further developed by enriching features and comparing it with other algorithms to enhance prediction accuracy.