Title: A study on severity of traffic accidents using road, weather and time characteristics

Authors: Zeynep Burcu Kizilkan; Ahmet Erdogan Asliyuce; Tugay Cengiz; Uğur Can Ersen

Addresses: Faculty of Management, Department of Industrial Engineering, Istanbul Technical University, 34367 Macka, Istanbul, Turkey ' Getir Perakende Lojistik A.Ş., 34337 Etiler, Istanbul, Turkey ' Hastavuk Gıda Tarım Hayvancılık Sanayi ve Ticaret A.Ş., 16130 Nilüfer, Bursa, Turkey ' Türkiye İş Bankası A.Ş., 34330 Levent, Istanbul, Turkey

Abstract: Mortality and severe injuries caused by traffic accidents are vital threats to society, therefore contributing factors to accidents are a major concern. Accident severity can be understood by attributes like human factors, the impact of road characteristics, weather, and accident time. Artificial neural networks (ANNs) are more practical and efficient to implement compared to other algorithms while computing risk levels using categorical data. Accordingly, ANNs are a well-researched and applied technique in traffic accident prediction models and determining contributing factors of traffic accidents. Previous research includes predominantly human impact. This paper aims to build a model to observe the impact of road, weather, and time characteristics rather than human factors on risk levels. Two models are constructed using ANNs, performance comparison indicates that both models reached a satisfactory certainty level. For further development, this model can be developed as a prevention system to enable the use of governmental institutions.

Keywords: accident severity; ANNs; artificial neural networks; traffic accident; machine learning; supervised learning; prevention system.

DOI: 10.1504/IJDS.2021.118948

International Journal of Data Science, 2021 Vol.6 No.2, pp.147 - 171

Received: 02 Jul 2020
Accepted: 30 Jul 2021
Published online: 12 Nov 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article