Predicting Number of Casualties During Accidents Using Machine Learning: Fid-Bau Portal

Predicting Number of Casualties During Accidents Using Machine Learning

Elawady, Ahmed / Alotaibi, Emran / Mostafa, Omar / Abuzwidah, Muamer

Over the years, there has been a growing interest in modeling the occurrence of road accidents. Researchers and policymakers have put in a lot of work to figure out why people get into accidents on the road and how to enhance highway safety. In most cases, accidents are associated with casualties in people. According to reports, over 1.35 million people die in road accidents each year, with 50 million people injured, resulting in an average of 3000 deaths and 30,000 injuries every day. Furthermore, the potential consequences have an economic and societal effect on healthcare expenses associated with accidents and disabilities. This research aims to use modeling approaches such as Support Vector Machine (SVM) and Artificial Neural Networks (ANN) to predict the number of casualties in accidents and compare these models' performance in terms of prediction accuracy. More than 91,000 accident records were obtained from London, including over 32 factors for each accident record. The results showed that the ANN models outperformed the SVM models in terms of Root Mean Square Error (RMSE) and training time with Narrow Neural Network (10 Neurons), achieving the best results. However, the opposite happens regarding Mean Absolute Error (MAE). Also, the variable importance results showed that the number of vehicles involved in the accidents is the most contributing factor that affects the number of casualties. This finding may be particularly beneficial to researchers and traffic authorities seeking effective and efficient accident management to enhance traffic safety.