The Use of XGBoost Algorithm to Analyse the Severity of Traffic Accident Victims

• Published in: Lontar komputer Vol. 14; no. 1; pp. 36 - 47
• Main Authors: Sukarsa, I Made, Dwi Rusjayanthi, Ni Kadek, Utami, Made Srinitha Millinia, Wisswani, Ni Wayan
• Format: Journal Article
• Language: English
• Published: Udayana University, Institute for Research and Community Services 27-10-2023
• ISSN: 2088-1541; 2541-5832
• DOI: 10.24843/LKJITI.2023.v14.i01.p04

Traffic accidents are still significant contributors to a fairly high death. Denpasar’s resort police record every traffic accident in the form of a daily report. The stored data can generate valuable information to improve policies and propagate better traffic practices. This research utilizes the classification technique with the XGBoost, random forest algorithm, and SMOTE method. The study shows that the SMOTE technique can increase the model's accuracy. Using the classification method with the two algorithms produces factors that affect the severity of traffic accident victims with feature importance. The feature importance obtained using the XGBoost model by counting the weight value for testing using the original dataset, the dataset for the type of two-wheeled vehicle, and the dataset of the kind of vehicle other than two-wheeled indicate that the variables influencing the severity of victims in road accidents are the time of accident between 00.00-06.00, the type of vehicle motorcycle, the type of opponent vehicle truck and pickup car, the age of the driver between 16-25, sub-district road status and front – side type of accident.