Kepatuhan Pembayaran Pajak Kendaraan Bermotor Menggunakan Algoritma Decision Tree Dan Random Forest Di Samsat Balige

Abstract

This study aims to analyze and predict the total category of Motor Vehicle Tax (PKB) payments based on payment attributes and vehicle types, which is important to improve the effectiveness of tax management and support more appropriate decision making in related agencies; within the theoretical framework, classification models such as Decision Tree and Random Forest are used to predict data categories by utilizing historical patterns in the dataset, because these algorithms are able to capture interactions between attributes and provide logical interpretations of the prediction results; the research methodology is carried out using secondary data of PKB payments for 2024 from Samsat Balige, which is divided into training data and test data for the classification process and its performance is evaluated using accuracy, precision, recall, and F1-Score metrics through the Performance operator in RapidMiner; the results of the study show that Random Forest produces a more balanced prediction distribution with 100% accuracy, while Decision Tree has 96% accuracy but tends to be biased towards the “Low” category, and analysis of important attributes such as Fines, Total Amount, and the number of Jeep and Truck type vehicles shows a significant influence on the PKB payment category; Thus, the research conclusion confirms that Random Forest is proven to be more effective and stable than Decision Tree in predicting the total PKB payment category, is able to capture complex patterns between attributes, and provides accurate predictions on relatively small datasets, making it the optimal choice for PKB data classification.