Pemodelan Resiko Kecelakaan Berbasis Kondisi Kendaraan dan Pengemudi

Abstract

Traffic accidents are particularly prone to occur mainly caused by vehicle speed, vehicle damage, alcohol influence, and fatigue. The study aims to model the risk of vehicle and driver-based accidents occurring across Queensland, Australia. The data in this study used a dataset of accident factors on Queensland state roads totaling 3412 accidents sourced from the Australian state government of Queensland. Research data period from 2001-2019. This research method uses multinomial logistic regression modeling analysis. The results of this study produced several models, namely; (1) Log odds in the risk level of death vs hospitalization will increase by 1,028 if affected by vehicle damage, increase by 0.731 if affected by fatigue, increase by 0.158 if affected by vehicle speed, increase by 0.151 if influenced by alcohol. (2) Log opportunities in the risk level of death vs. medical care will increase by 0.786 if affected by vehicle damage, increase by 0.375 if affected by fatigue, decrease by 0.003 if affected by vehicle speed, decrease by 0.078 if influenced by alcohol. (3) Log odds in the risk of death vs minor injury will increase by 0.484 if affected by vehicle damage, increase by 0.245 if affected by fatigue, decrease by 0.156 if affected by vehicle speed, decrease by 0.266 if influenced by alcohol. (4) Log odds in the risk of death vs property damage will increase by 1,254 if affected by vehicle damage, increase by 0.828 if affected by fatigue, increase by 0.185 if influenced by vehicle speed, increase by 0.128 if influenced by alcohol. The validation test value with crosstab method explains that the accuracy result of level 1 has an accuracy value of 0.99 and inaccuracy of 0.01 then the result of level 2 to level 5 has an accuracy value of 1.