<![CDATA[MyTelkomsel is a customer service application developed by one of the largest cellular operators, with more than 100 million users. Due to the high volume of application users, sentiment analysis is essential for examining user opinions to optimize service quality. However, sentiment classification often faces challenges caused by imbalanced sentiment class distributions, which can affect model performance. This study analyzes sentiment toward the MyTelkomsel application using the Support Vector Machine (SVM) algorithm, focusing on evaluating the performance of Linear, RBF, and Polynomial kernels. The dataset consisted of 1,000 user reviews randomly collected from the Google Play Store, with positive and negative labels assigned based on the Indonesia Sentiment Lexicon (InSet). The dataset was divided into training and testing sets using an 80:20 ratio. The model development process was carried out using RapidMiner. The optimal performance was achieved by the Linear kernel through the implementation of the Synthetic Minority Over-sampling Technique (SMOTE) and K-Fold Cross Validation, resulting in an accuracy of 100%, precision of 100%, recall of 100%, and F1-score of 100%. These results indicate that the data can be effectively separated using a linear boundary. SMOTE was applied to address class imbalance in the dataset, while K-Fold Cross Validation (k = 10) was used to ensure the absence of overfitting by testing the entire dataset divided into 10 folds. The findings of this study can serve as a foundation for optimizing application services, enabling improvement strategies to be implemented in accordance with feedback derived from user reviews.]]>
