Sentiment Analysis of Transportation Application Reviews with SVM on Handling Imbalanced Data Using SMOTE

Transportation has become an inseparable element of Indonesian society. In April 2024, the number of motor vehicles in Indonesia reached 161,787,250 units, causing traffic congestion in various regions, especially in large cities. The Mitra Darat application, as one of the transportation company applications, was initially known as "Teman Bus." This application has evolved into a multi-service platform such as BRT Nusantara, KSPN, and Perintis. This study analyzes sentiment toward reviews of transportation company applications to improve public transportation services in Indonesia. Sentiment analysis has been carefully conducted on user reviews in the Google Play Store using the Support Vector Machine (SVM) technique. The imbalanced data was addressed using the SMOTE and SMOTE-ENN techniques. The model was evaluated using accuracy, precision, recall, and F1-Score metrics. The results show that the accuracy achieved by the Support Vector Machine (SVM) with SMOTE is 85.68%, precision is 92.35%, recall is 84.12%, and F1-Score is 88.04%, demonstrating its effectiveness in sentiment classification. The use of SMOTE-ENN increased precision to 97.33% but decreased recall and accuracy to 80.80%. These findings affirm the superiority of Support Vector Machine (SVM) with SMOTE in handling imbalanced data and provide recommendations for improving the adoption of public transportation in Indonesia.