Analisis Kepuasan Masyarakat Terhadap Kinerja Bupati Labuhanbatu Selatan Periode 2021-2024 Menggunakan Metode Decision Tree dan Naive Bayes

Abstract

This study was conducted to analyze the level of customer satisfaction with services by comparing the performance of two classification methods, namely Decision Tree and Naive Bayes, so that an accurate model can be obtained to assist decision making. This problem is important because understanding customer satisfaction patterns can be a strategic basis in improving service quality and maintaining loyalty. The theoretical basis used refers to the concept of machine learning classification, where Decision Tree forms a branching rule-based model based on attributes, while Naive Bayes relies on probability calculations based on Bayes' theorem with the assumption of independence between features. The research methodology includes data collection stages, pre-processing to ensure data quality, model training with both methods, and performance evaluation using Test & Score and Confusion Matrix. Based on the classification results, the Decision Tree method produces fairly good accuracy, precision, and recall, but the Naive Bayes method shows higher performance with an accuracy of 91.67%, a precision of the "Satisfied" class of 98.11%, and a recall of 92.86%, which indicates a very good level of prediction accuracy especially for the majority class. Evaluation of both methods shows that Naive Bayes excels in capturing existing data patterns, although Decision Tree still has good interpretability for classification rule analysis. In conclusion, both methods are capable of classifying customer satisfaction data with adequate performance, but Naive Bayes is recommended as the primary model due to its higher and more consistent evaluation results, while Decision Tree can be used as an alternative when model interpretation is a priority.