A Stochastic Delay Prediction Model for Real-Time Incident Management

TL;DRAbstract

Incident management policies always involve making real-time decisions under uncertain and rapidly changing conditions. To support this process, a number of computer models have been developed to predict incident severity, although few of these can truly accommodate the uncertainty that exists in this process. In this paper, we demonstrate that failing to account for this consistently and systematically underestimates the impact of incidents, potentially to a large enough degree that faulty incident management decisions are made (simulation results indicate underestimation on the order of 20-50%). As a result, we develop a new delay prediction model that explicitly provides predictions in the context of uncertain incident duration, which eliminates this source of error. Analytical incident delay formulae are extended to account for uncertain incident duration, and simulation with Monte Carlo sampling is undertaken to study scenarios which are too complicated for exact analysis. These i

Chat with Paper

AI Agents for this Paper