Intelligent nanoscopic road safety model for cycling infrastructure

This paper is concerned with the development of an intelligent nanoscopic safety model for cycling safety. The present models are primarily focused on motorists modelling at an aggregate level. In this work, a framework for safety analysis is proposed consisting of a) Data collection unit, b) Data storage unit, and c) Knowledge processing unit. The predictive safety model is developed in the knowledge-processing unit using supervised deep learning with neural network classifier and gradient descent backpropagation error function. This framework is applied to a case study in Tyne and Wear County in England's northeast using the crash database. An accurate safety model (88% accuracy) is developed with the output of the riskiest age and gender group, based upon the specific input variables. The most critical variables affecting the safety of an individual belonging to a particular age and gender groups are the journey purpose, traffic flow regime and variable environmental conditions it is subjected to. It is hoped that the proposed framework can help in better understanding of cycling safety, aid the transportation professional for the design and planning of intelligent road infrastructure network for the cyclists.