Citation

Niyogisubizo J, Liao L, Sun Q, Nziyumva E, Wang Y, Luo L, Lai S, Murwanashyaka E. Int. J. Intell. Transp. Syst. Res. 2023; 21(1): 240-258.

Copyright

(Copyright © 2023, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s13177-023-00351-7

PMID

unavailable

Abstract

Smart cities came out as highly knowledgeable bio-networks, offering intelligent services and innovative solutions to urban problems. With rapid development, urbanization, and population pressure, traffic congestion and collisions are increasing substantially in road highway zones. Recently, traffic collisions have become one of the hugest national health problems in many cities of the world. Hence, crash injury severity prediction is vital for informing responsible authorities and the public to find alternative ways of dealing with its adverse effects, accordingly improving traffic safety and reducing traffic congestion. In predicting crash injury severity, researchers have explored and applied several techniques to aid in traffic injury management. However, the performance of many techniques suffers from some inherent limitations, including overgeneralization, lack of interpretability for humans, and low-performance accuracy. To address these issues, this paper proposes a novel computational framework based on improved wide and deep learning methods to predict accurately crash injury severity in the context of smart cities. On the crash dataset of New Zealand cities from 2000 to 2020, the proposed model has demonstrated better performance in comparison with the benchmark algorithms. Moreover, SHAP (SHapley Additive exPlanation) is employed to interpret the results and analyze the importance of each determinant of crash severity. The proposed method can help to prevent traffic crashes in smart cities and take proactive measures before the occurrence, also trauma centers can refer to this information to dispatch proper emergency service equipment swiftly and assist injuries to get direct medical care regardless of the crash location.

Language: en

Keywords

Crash injury severity; Deep learning; Machine learning; SHAP; Smart cities; Traffic accident prediction