Comparison of SVM and KNN Methods for the Integratin of MyIndiHome into MyTelkomsel Application

Abstract

This study aims to analyze user sentiment toward the merger of the MyIndiHome application into the MyTelkomsel platform conducted by PT Telkom Indonesia. In the digital era, the integration of these two customer service applications represents a strategic step to create a unified digital ecosystem. However, this merger has also generated diverse user responses, reflected in various reviews on the Google Play Store. To analyze these opinions, 1,556 user reviews were collected using the web scraping technique. The preprocessing stage included cleaning, tokenizing, filtering, normalization, stemming, and the application of the Synthetic Minority Over-Sampling Technique (SMOTE) to address class imbalance. Two machine learning algorithms, Support Vector Machine (SVM) and K-Nearest Neighbor (KNN), were applied to classify sentiments into positive, negative, and neutral categories. The experimental results showed that SVM achieved higher accuracy (86.2% before SMOTE and 84.9% after SMOTE) compared to KNN (83.7% before SMOTE and 67.6% after SMOTE). These results indicate that SVM performs more effectively and consistently in handling high-dimensional text data than KNN. Therefore, SVM is considered a more reliable algorithm for sentiment classification in this context. The findings provide valuable insights for PT Telkom Indonesia in understanding user perceptions, improving service quality, and enhancing user experience following the digital integration of MyIndiHome into MyTelkomsel.