<![CDATA[The selection of a suitable classification model is important in text-based sentiment analysis, especially in conditions of unbalanced data distribution. Naive Bayes and Support Vector Machine (SVM) are two algorithms that are often used in classification, but the comparison of their performance on unbalanced data still needs to be further reviewed. This study aims to compare the performance of the two algorithms in classifying public sentiment towards the Indonesia Smart Card (KIP) Lecture Program. The implementation of the KIP Lecture Program still faces challenges in the accuracy of aid distribution. This situation raises discussions and various controversies among the public, especially on the X platform. The data used were 1,644 tweets, with a distribution of negative sentiment of 1,392 tweets and positive tweets of 252. To overcome the imbalance of data class distribution, the Synthetic Minority Oversampling Technique (SMOTE) method is used. Based on the evaluation results, before SMOTE was applied, SVM obtained 92% accuracy and 91% precision, 77% recall, while Naive Bayes obtained 79% accuracy, 68% precision, and 78% recall. After the application of SMOTE, SVM performance significantly improved with accuracy, precision, and recall reaching 99%, while Naive Bayes improved to 95% on all metrics. These results show that although SVM excels in higher accuracy, Naive Bayes shows a more stable performance on the data neither after nor after the balancing process is performed.]]>
