<![CDATA[The increasing number of motorized vehicle users in Grobogan Regency has resulted in an increasing number of traffic violations that have the potential to cause accidents. Each violation has a different level of severity, so a classification method is needed to assist in determining the appropriate action. This study aims to compare the performance of the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) methods in classifying the severity of traffic violations into three categories: minor, moderate, and severe. The research data was obtained from the Grobogan Police Department in 2020–2024 with a total of 38,318 cases. The research stages include data preprocessing, transformation, labeling using the K-Means method, data division, feature selection with Chi-Square, and classification using KNN and SVM. The evaluation results show that the KNN method produces an accuracy of 87.7% with an F1-score of 84.8%, but the AUC (64.3%) and MCC (21.1%) values are still low, making it less than optimal in distinguishing classes. Meanwhile, the SVM method excelled in all evaluation metrics, with accuracy, precision, recall, F1-score, and MCC values each reaching 98.5%. The distribution of classification results showed that KNN tended toward the mild class, while SVM was able to separate the three classes more evenly. Thus, SVM proved more effective in classifying traffic violation severity in Grobogan, producing more accurate results.]]>
Copyrights © 2026
