Using artificial intelligence to derive a public transit risk index

Abstract

A terrorist attack on the public transportation system of a city can cripple its economy. Uninformed investments in countermeasures may result in a waste of resources if the risk is negligible. However, risks are difficult to quantify in an objective manner because of uncertainties, speculations, and subjective assumptions. This study contributes a probabilistic model, validated by ten different machine learning methods applied to the fusion of six heterogeneous datasets, to objectively quantify risks at different jurisdictional scales. The risk index is purposefully simple to quickly inform a proportional prioritization of resources to make fair investment decisions that stakeholders can easily understand, and to guide policy formulation. The main finding is that the risk indices among public transit jurisdictions in the United States distribute normally. This result enables agencies to evaluate the quality of their risk index calculations by detecting an outlier or a large deviation from the expected value.