Kinerja Naive Bayes dan SVM pada Data Survei Tidak Seimbang: Studi Klasifikasi Kepuasan Masyarakat

The utilization of Public Satisfaction Survey (SKM) data has not been optimal, highlighting the need for an effective classification method to determine the level of public satisfaction. This study aims to classify satisfaction levels using the 2024 SKM data from the Regional Civil Service and Training Agency (BKPPD) of Grobogan Regency, employing Naive Bayes and Support Vector Machine (SVM) algorithms. This quantitative research uses nine service elements rated on a scale of 1 to 4 as features, with satisfaction level as the target variable. The dataset consists of 303 entries: 156 “very satisfied,” 115 “satisfied,” and 32 “dissatisfied.” Random oversampling was applied to address class imbalance. Model performance was evaluated using accuracy, precision, recall, and F1-score, both before and after oversampling. Results showed Naive Bayes achieved 96.72% accuracy, while SVM scored 95.08%. After oversampling, SVM accuracy significantly improved to 98.36%, while Naive Bayes slightly decreased to 95.08%. Precision, recall, and F1-scores also demonstrated strong performance across all classes. This study is expected to support the improvement of public service delivery at BKPPD Grobogan and similar institutions.