Redes neuronales artificiales aplicadas en sistemas de predicción para la seguridad vial

Abstract

Abstract: The present investigation show a prediction model of traffic accidents in two road axles at South of Ecuador, in the Loja province. The model was developed use Artificial Neural Networks (ANN) with the goal, to identify its suitability for the prediction of traffic accidents, under conditions such as: curvature radios and slope of the tracks. The proposed model show a multilayer network architecture (feedforward), which has three layers (entry, hidden and exit), the number of entry and hidden neurons vary according to the model. The training function for to the modeling is the Bayesian regulation backpropagation. The training phase and validation of the models were made use probabilistic mistakes calculations.