Ulasan pengguna aplikasi Green SM pada Google Play Store berjumlah besar dan berbentuk data teks tidak terstruktur sehingga sulit dianalisis secara manual. Penelitian ini bertujuan mengklasifikasikan sentimen ulasan pengguna ke dalam tiga kelas, yaitu positif, negatif, dan netral, menggunakan algoritma Support Vector Machine (SVM) dengan ekstraksi fitur Term Frequency–Inverse Document Frequency (TF-IDF) berdasarkan kerangka kerja Cross-Industry Standard Process for Data Mining (CRISP-DM). Data diperoleh melalui proses web scraping sebanyak 1.087 ulasan, kemudian dilakukan text preprocessing dan pelabelan sentimen menggunakan metode lexicon-based yang divalidasi secara manual sehingga diperoleh dataset akhir sebanyak 1.032 ulasan, terdiri atas 542 ulasan negatif, 434 ulasan positif, dan 56 ulasan netral. Hasil penelitian menunjukkan bahwa model SVM terbaik menggunakan kernel Radial Basis Function (RBF) dengan parameter C = 10 dan gamma = scale, yang menghasilkan nilai accuracy sebesar 85,51%, precision sebesar 85,96%, recall sebesar 85,51%, dan F1-score sebesar 84,70%. Selain itu, analisis frekuensi kata menunjukkan bahwa kata "driver" dan "sopir" merupakan istilah yang paling sering muncul dalam ulasan pengguna, sehingga aspek pelayanan pengemudi menjadi perhatian utama pengguna aplikasi Green SM. Hasil penelitian ini diharapkan dapat menjadi bahan evaluasi bagi perusahaan dalam meningkatkan kualitas aplikasi dan layanan yang diberikan.

User reviews of the Green SM application on Google Play Store are available in large numbers and consist of unstructured text, making manual analysis inefficient. This study aims to classify user review sentiment into three categories—positive, negative, and neutral—using the Support Vector Machine (SVM) algorithm with Term Frequency–Inverse Document Frequency (TF-IDF) feature extraction based on the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework. A total of 1,087 reviews were collected through web scraping, followed by text preprocessing and lexicon-based sentiment labeling with manual validation, resulting in a final dataset of 1,032 reviews consisting of 542 negative, 434 positive, and 56 neutral reviews. The best-performing SVM model employed the Radial Basis Function (RBF) kernel with C = 10 and gamma = scale, achieving an accuracy of 85.51%, precision of 85.96%, recall of 85.51%, and an F1-score of 84.70%. In addition, word frequency analysis revealed that the terms "driver" and "sopir" (driver) appeared most frequently in user reviews, indicating that driver service is the primary concern among Green SM users. The findings of this study are expected to provide useful insights for the company in improving the quality of its application and services.