Identifying Path Diversions of Hazardous Materials Vehicles for Security Alerts

Abstract

The transport of hazardous materials (hazmats) poses a threat to society and the environment. Beyond the issues pertaining to unintentional events, the potential exists for hazmat vehicles to be used for malicious purposes. A key to combating this threat is the detection of vehicle deviations from their normal paths. This article presents a probabilistic neural network approach to identifying path deviations and classifying the threat level at each node in the network. The methodology is illustrated on the network between Baltimore, Maryland, and Washington, D.C. The selected model obtained an accuracy of 93.51%, which was not the highest of all the models. Selection should depend not only on the overall error rate but also on the error severity. One-category errors (e.g., classifying dangerous nodes as suspicious) were less severe than two-category errors (e.g., classifying dangerous nodes as safe). False positives, which require a response from a system monitor or law enforcement, were considered as well as false negatives, which do not send warnings when they should. The selected neural network would require more human involvement but would provide a more secure transportation network.