Analisis Sentimen Ulasan Aplikasi Gojek Menggunakan Support Vector Machine

Muhammad Fariz (Universitas Muhammadiyah Prof. Dr. HAMKA)
Ade Davy Wiranata (Universitas Muhammadiyah Prof. DR. HAMKA)

Progresif: Jurnal Ilmiah Komputer
Vol. 22 No. 3 (2026): Juli

Evaluating Gojek’s service quality can be achieved by extracting valuable insights from the Google Play Store review corpus. However, the surging volume of data necessitates computational automation to substitute inefficient manual analysis. This study develops a sentiment analysis model using the Support Vector Machine (SVM) algorithm applied to 2,000 scraped reviews. The corpus undergoes comprehensive text preprocessing, rating-based polarity labeling, and lexical feature extraction via Term Frequency-Inverse Document Frequency (TF-IDF). Model validation demonstrates an impressive predictive performance, achieving an accuracy of 86.21%. The distribution of public opinion reveals a dominance of positive sentiment at 57.47%, outperforming the negative cluster at 42.53%. Empirically, the methodological fusion of TF-IDF and SVM is highly reliable as an analytical instrument to dissect consumer perceptions for future service improvements. Keywords: Sentiment Classification; Support Vector Machine; TF-IDF Weighting; Service Evaluation; Gojek. Abstrak Evaluasi kualitas layanan Gojek dapat diekstraksi dari korpus ulasan Google Play Store. Namun, lonjakan volume data menuntut otomatisasi komputasional guna mensubstitusi analisis manual yang inefisien. Penelitian ini mengonstruksi model analisis sentimen menggunakan algoritma Support Vector Machine (SVM) terhadap 2.000 entri ulasan hasil web scraping. Korpus diproses melalui tahapan prapemrosesan teks komprehensif, pelabelan polaritas berbasis rating, serta ekstraksi fitur leksikal mendayagunakan metode Term Frequency-Inverse Document Frequency (TF-IDF). Validasi arsitektur klasifikasi membuktikan performa prediktif yang impresif, dengan raihan akurasi menyentuh 86,21%. Peta distribusi opini publik mendemonstrasikan dominasi sentimen afirmatif (positif) sebesar 57,47%, mengungguli klaster negatif di angka 42,53%. Secara empiris, fusi metodologis antara pembobotan TF-IDF dan SVM terbukti sangat andal sebagai instrumen analitik komputasional guna membedah persepsi konsumen demi mendukung evaluasi perbaikan layanan Gojek di masa depan.