Perkembangan teknologi informasi dan komunikasi, khususnya dalam aplikasi mobile di platform Android, telah membawa perubahan signifikan dalam cara pengguna berinteraksi dan memberikan umpan balik. Dalam peelitian ini, analisis sentimen pada ulasan pengguna menjadi sangat penting untuk memahami kepuasan dan kebutuhan pengguna. Penelitian ini bertujuan mengembangkan model klasifikasi sentimen untuk ulasan aplikasi X (Twitter) di Google Play Store menggunakan metode Support Vector Machine (SVM). Sebanyak 1.194 ulasan dikumpulkan dan diproses melalui tahap preprocessing yang mencakup pembersihan data, tokenisasi, penghapusan stop words, dan konversi teks menggunakan TF-IDF (Term Frequency-Inverse Document Frequency). Data dibagi dengan rasio 80:20, di mana 80% digunakan untuk pelatihan dan 20% untuk pengujian. Hasil evaluasi model menunjukkan akurasi keseluruhan sebesar 90%. Model SVM menunjukkan performa sangat baik pada kelas positif (0) dan netral (2), dengan precision masing-masing 89% dan 90%, serta recall masing- masing 95% dan 96%. Namun, model mengalami kesulitan dalam mengidentifikasi kelas netral (1), dengan precision 100% tetapi recall hanya 17%, menghasilkan F1-score rendah sebesar 0.29. Macro average menunjukkan keseimbangan dalam precision (93%) tetapi kurang memadai dalam recall (69%) dan F1-score (71%). Weighted average menunjukkan akurasi keseluruhan model dengan precision, recall, dan F1-score masing-masing 90%, 90%, dan 88%. Temuan ini menegaskan efektivitas model SVM dalam mengklasifikasikan sentimen positif dan netral, namun menunjukkan tantangan signifikan dalam klasifikasi sentimen negatif.

The advancement of information and communication technology, particularly in mobile applications on the Android platform, has brought significant changes in how users interact and provide feedback. In this research, sentiment analysis of user reviews has become crucial for understanding user satisfaction and needs. This study aims to develop a sentiment classification model for reviews of the X (Twitter) app on the Google Play Store using the Support Vector Machine (SVM) method. A total of 1,194 reviews were collected and processed through preprocessing stages that included data cleaning, tokenization, stop words removal, and text conversion using TF-IDF (Term Frequency-Inverse Document Frequency). The data was split with an 80:20 ratio, where 80% was used for training and 20% for testing. The model evaluation results indicate an overall accuracy of 90%. The SVM model demonstrated excellent performance on the positive (0) and neutral (2) classes, with precision of 89% and 90%, and recall of 95% and 96%, respectively. However, the model faced difficulties in identifying the neutral (1) class, with a precision of 100% but a recall of only 17%, resulting in a low F1- score of 0.29. The macro average showed a balance in precision (93%) but was inadequate in recall (69%) and F1-score (71%). The weighted average maintained an overall model accuracy with precision, recall, and F1-score of 90%, 90%, and 88%, respectively. These findings highlight the effectiveness of the SVM model in classifying positive and neutral sentiments but indicate significant challenges in classifying negative sentiments.