Prediction of traffic accidents based on weather conditions in Gilan province using artificial neural network

Abstract

Introduction: Road traffic accidents are one of the leading causes of death worldwide, including Iran. There are several factors involved in the occurrenceof them; using different models, these factors can be identified and theoccurrence of road traffic accidents can be predicted. The purpose of this studywas to predict road traffic accidents based on weather conditions usingartificial neural network model.Methods: In the present study, traffic data during the years 2014 to 2017,were examined using a multilayer perceptron network. Network input variablesincluded minimum temperature, average temperature, average rainfall,maximum wind speed, glaciation, air pressure, fog concentration and outputvariable was the number of accidents.Results: The designed network with seven neurons in the input layer, fourneurons in the middle layer, and one neuron in the output layer with Lunberg-Marquardt optimization function and sigmoid tangent transfer function in themiddle layer and linear transmission function in the output layer was selectedas the optimal network. The results showed that the designed network with thecorrelation coefficient of 0.90 and mean square error of 0.01 has a high abilityto predict road traffic accidents.Conclusion: The results showed that the artificial neural network has a goodperformance for predicting road traffic accidents. Given the importance ofpredicting road traffic accidents and its role in promoting the health of peoplein such accidents, the results of this study can be used to expand more effectivepreventive measures for policy makers and researchers