ANALISIS SENTIMEN PENGGUNA APLIKASI LAYANAN PUBLIK BERBASIS WEB MENGGUNAKAN KOMPARASI ALGORITMA SUPPORT VECTOR MACHINE DAN NAIVE BAYES

Abstract

Evaluating public service application feedback requires rapid and objective automated analysis. This study presents a comparative evaluation of Support Vector Machine (SVM) and Multinomial Naive Bayes (MNB) classifiers for categorizing public feedback on web-based government services into positive and negative sentiments. The corpus of 1,000 Indonesian textual reviews underwent tokenization, stopword removal, and Nazief-Adriani stemming alongside TF-IDF feature extraction. Experimental results confirm that linear kernel SVM achieved superior classification accuracy of 89.50% and an 89.20% F1-Score, outperforming Naive Bayes (84.00%). The findings deliver quantitative insights for public administrative bodies to address recurring operational grievances.