ANALISIS SENTIMEN PEMBELIAN BAHAN BAKAR MINYAK PADA APLIKASI MyPertamina DENGAN METODE NAIVE BAYES CLASSIFIER DAN SYNTHETIC MINORITY OVERSAMPLING TECHNIQUE

The implementation of the MyPertamina application policy for subsidized fuel purchases has received various responses from the public, expressed through social media. These responses can be classified into neutral, positive, and negative feedback. Manual analysis can be time-consuming, so the Naive Bayes Classifier (NBC) method can be used for quick and accurate sentiment analysis of public responses to the implementation of the MyPertamina application for subsidized fuel purchases. The research aims to analyze sentiment using the NBC method and implement the Synthetic Minority Oversampling Technique (SMOTE) on the application of MyPertamina for subsidized fuel purchases in the community. The dataset in this study is divided into three ratios: 30% for testing set, 40%, and 50%. The sentiment analysis results using the NBC and SMOTE classification methods with a 30% training set ratio show the best outcome. Initially, there were 972 data points, which were preprocessed to 712, and then the SMOTE algorithm was implemented to balance the training set. The results showed that 38% were neutral responses, 35% were positive, and 27% were negative, with an accuracy of 84%.