Utilizing machine learning models to predict the severity of road accidents: A case study on the US Accidents Dataset from ۲۰۱۶ to ۲۰۲۳

Road transportation is the primary method of travel in the United States, making road safety a significant concern. Every year, numerous accidents lead to significant loss of life and damage to road infrastructure. Previous studies frequently encounter constraints, including limited datasets restricted to specific geographic areas, significant dependence on intricate data, and difficulties implementing findings in real time. In order to address these limitations, this research presents an innovative approach using machine learning algorithms for accurately forecasting traffic accidents in real time by utilizing easily accessible data. This study revolves around the Kaggle dataset 'US-Accidents,' which contains more than ۷ million records gathered from multiple sources. The findings of this study showed the optimal efficiency of machine learning algorithms in predicting the severity of road accidents. Gradient Boosting (accuracy of ۹۱.۸۵%) emerged as the best-performing model, with exceptional accuracy. Random Forest (accuracy of ۹۱.۷۹%) and Decision Tree (accuracy of ۸۹.۷۵%) models demonstrated strong performance, suggesting their appropriateness for intricate data analysis. Multi-Layer Perceptron (MLP), with an accuracy of ۸۵.۹۵%, and Support Vector Machines (SVMs), with an accuracy of ۸۴.۶۵% models, had results close to each other. At the same time, the MLP performed less than expected. Logistic Regression (accuracy of ۸۰.۲۷%), which served as the baseline, showed the lowest level of accuracy.

کلیدواژه ها:

Accident Severity ، Machine Learning ، Random Forests ، Decision Trees ، Support Vector Machines (SVMs) ، Logistic Regression ، Gradient Boosting ، Multi-Layer Perceptron (MLP)