Highway Traffic Accidents Prediction Using Artificial Neural Network: A Case Study of Freeways in Spain
Highway Traffic Accidents Prediction Using Artificial Neural Network: A Case Study of Freeways in Spain
-
DOI: https://doi.org/10.22533/at.ed.13174124090110
-
Palavras-chave: Redes Neurais Artificiais RNAs, Acidentes de Trânsito
-
Keywords: Artificial Neural Networks ANNs, Traffic Accidents
-
Abstract: In recent years, Spain has witnessed a significant reduction in the accident rate, attributable to the improved behavior of road users. However, there remains a pressing need for enhancements in various areas. Notably, 2016 marked the first time in 13 years that the number of deaths increased by 7% compared to the previous year. This paper undertakes an analysis and prediction of road traffic accidents (RTAs) at severe accident locations on Spanish highways, employing Artificial Neural Networks (ANNs) with a Feedforward learning algorithm. This approach serves as a valuable decision-making tool for policymakers in infrastructure management, contributing to advancements in transportation safety research. The ANN, a potent technique with a track record of success in analyzing historical data to forecast future trends, is explored to predict the number of highway accidents in Spain. The paper proposes a method to select the most effective ANN model using accident data spanning from 2014 to 2017. The model incorporates variables such as highway sections, year, section length (km), annual average daily traffic (AADT), average horizontal curve radius, degree of vertical curvature, and traffic accidents with the percentage of heavy vehicles. In the development of the ANN model, the sigmoid activation function is employed in conjunction with the Levenberg-Marquardt algorithm, incorporating varying numbers of neurons. The results of the model indicate that the estimated traffic accidents, based on appropriate data, closely align with actual traffic accidents, making them suitable for forecasting traffic accidents in Spain. This underscores the potential of ANNs as a robust tool for analyzing and predicting traffic accidents and casualties.
- Ali Alqatawna