Comparative Sentiment Analysis of Provider X Application Reviews Using Support Vector Machine, Random Forest, and Naïve Bayes Algorithms Based on TF-IDF and SMOTE

Hamama Kamtelat
Nirwan Moningka
Agung K Henaulu (Universitas Darussalam Ambon)
Haris Kolengsusu
Rahul Lestaluhu
Muhamad Alvuad Mualo

SITEKIN: Jurnal Sains, Teknologi dan Industri
Vol. 23 No. 2 (2026): June 2026

In the era of digital transformation, the Provider x application has become the main digital touchpoint that reflects corporate credibility through customer experience. User reviews on the Google Play Store are an authentic business intelligence asset, but their massive volume requires machine learning solutions for objective and measurable analysis. This research aims to conduct comparative sentiment analysis by applying three labeling categories (positive, negative and neutral) to capture user opinions more comprehensively. Using a dataset of 10,983 clean reviews, this research applies rigorous text pre-processing and feature extraction using TF-IDF. To overcome class imbalance, the SMOTE (Synthetic Minority Over-sampling Technique) technique is integrated into the model. This research compares three main algorithms: Support Vector Machine (SVM), Random Forest, and Naïve Bayes. Experimental results show that Random Forest excels as the best model with the highest accuracy rate of 82%, significantly surpassing SVM (73%) and Naïve Bayes (71%). These findings prove the effectiveness of ensemble structures in processing high-dimensional text features and provide empirical insights for developers to improve services based on customer voice.