<![CDATA[Traffic accidents have been a serious problem caused by various factors such as road conditions, driver behavior, and weather. To understand the pattern of victim severity, a classification approach capable of handling imbalanced data and irrelevant features was needed. This study aimed to classify the status of accident victims using the Neighbor Weighted K-Nearest Neighbor (NW-KNN) method, equipped with backward elimination for feature selection. Backward elimination was employed to reduce insignificant features and improve accuracy.The case study for this research involved the status of accident victims in Bontang City, with a sample size of 93 cases. There were nine features in this study: accident victim status, accident time, road density, road function, road surface condition, speed limit at the location, road slope, and road status.The research results showed that the best parameter combination for classification using the NW-KNN method with backward elimination was K = 7 and E = 3. The "type of accident" feature was eliminated, leaving 8 features. Classification results using the NW-KNN method with backward elimination yielded an accuracy of 88.89%, demonstrating an improvement in classification performance for identifying the status of traffic accident victims. Thus, this method proved to be an effective approach for traffic accident analysis in Bontang City.]]>
Copyrights © 2025
