<![CDATA[Traffic accidents are an important factor in improving public safety in West Java Province, because the population movement rate there is very high. The high number of accidents is directly related to increased deaths and material losses, but the use of historical data is still limited to administrative archiving tasks without any process of identifying regional vulnerability patterns. This study aims to classify accident-prone areas using Data Mining techniques and the K-Means Clustering algorithm, as well as applying the CRISP-DM framework approach. The analyzed dataset comes from the Indonesian National Police Traffic Accident Database (Pusiknas Polri) for the period 2020 to 2025, consisting of 552 observations with indicator variables covering the number of fatalities, serious injuries, and minor injuries. The determination of the most appropriate number of clusters was tested using the Silhouette method to ensure more accurate and objective modeling results. The analysis shows that the number of clusters (k=3) is the most appropriate, with a Silhouette metric value of 0.398. The application of the model produces three levels of risk: the red zone, which indicates high risk with 51 cases and the highest mortality rate; the yellow zone, which indicates moderate risk with 189 cases; and the green zone, which indicates low risk with 312 cases. The visualization of these mapping results is expected to be an important tool for the police and local governments in formulating mitigation policies, improving patrol efficiency, and accelerating infrastructure improvements in high-risk areas, thereby reducing the number of accidents in the future.]]>
Copyrights © 2026
