Development of a Safety System for Intelligent Cyclist modelling

This paper is concerned with the modelling of cyclist road traffic crashes by considering multiple factors affecting the safety of cyclists. There are very few works in the literature dealing with such a problem. The available models in the literature are only based upon the probabilistic function of human error. In this study, we propose an intelligent safety system for modelling cycling infrastructure. The historic crash dataset for the Tyne and Wear County, north-east of England is used as a case study. There are five predictive road safety models develops using the Artificial Neural Network, with the output for the riskiest road type infrastructure. The study demonstrates that infrastructure, spatial variables, personal characteristics, and environmental conditions affect safety, which can also be used for predicting safety. These identified variables are modelled both individually and in combination with each other, and a plausible high accuracy is achieved in all the five models (> 85% accuracy). This demonstrates the benefit of using ANN for effective and efficient modelling of the safety variable for infrastructure design and planning. It is hoped that the proposed model can help in designing better cyclist infrastructure and contribute towards the development of a sustainable transportation system.