Work

OA20111. Crash injury severity prediction for passenger cars with machine learning algorithms

Year: 2025

Type: article

Abstract: Background Road traffic crashes represent a major public health concern, so it is of significant importance to understand the factors associated with the increase in injury severity of its interveners when involved in a road crash. Determining such factors is essential to help decision-making in road safety management, improving road safety, and reducing the severity of future crashes. Artificial Intelligence is probably the most important topic of research nowadays. Machine learning is a very prominent and successful approach in modern AI and is used in the work to predict crash injury severity. Methods For this study, all police-reported road accidents involving at least one injured passenger car occupant that occurred in Portugal over twelve years, from 2010 to 2022, were considered. The information present in the data (a total of 117 variables) can be divided into 3 parts: accident information (50 variables), occupants’ information (43 variables), and a third part related to the other driver involved (24 variables), when this exists. The size of the crash database is very large containing about 6000 fatalities, 20000 severe injuries and 600000 slight injuries. A full exploratory analysis of the data was performed for all variables in the dataset. To begin with, the exploratory analysis was performed on the original dataset. Afterwards, the continuous variables were discretized, and considering the previous exploration analysis, some classes were grouped within the different variables. A new exploration analysis was performed on this new discretized and regrouped dataset. Results Different machine learning methods are investigated and discussed in this study, namely Decision Tree (DT), Ordered Logistic Regression (LR), Random Forest (RF), Gradient Boosting (GB), Extreme Gradient Boosting (XGB), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM) . To compare the models’ performance the following metrics were used: sensitivity, specificity, precision. Key messages • AI with machine learning methods provides a suitable framework to build a proper predictive model, allowing researchers and practitioners to evaluate more accurately the risk factors of crashes. • Car accidents account for the majority of road victims in the world, and tools that can predict injuries are very important for taking effective prevention measures. Topic Injury prediction, Machine Learning, Traffic crashes.

Source: European Journal of Public Health

DOI: 10.1093/eurpub/ckaf165.111

Authors: J Dias, G Araújo, C Amado

Institutions: Instituto Superior Técnico, Centro de Matemática Computacional e Estocástica

Language: English

Topic: Traffic and Road Safety

Subfield: Safety, Risk, Reliability and Quality

Field: Engineering

Domain: Physical Sciences

SDGs: Good health and well-being, Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $3,181