Skip to main content
Infrastructure Resilience Conference 2018

Full Program »

Quantifying Community Resilience Using Hierarchical Bayesian Methods: A Case Study on Recovery from Power Outages

Resilience is often referred to as the ability to recover from the occurrence of a disruptive event. Various metrics have been developed to measure resilience for different types of hazards, but a common issue in quantifying those metrics and modeling them using data-driven methods is the lack of recovery data from historical events. Sparse data result in challenges in measuring and predicting resilience metrics within an acceptable level of accuracy. In this paper, the rate of recovery to the level prior to the occurrence of a disruptive event is used as the metric for resilience. Three statistical models are considered, hierarchical Bayesian model, Bayesian kernel model, and multivariate Poisson generalized linear model, to model the recovery rate with sparse data. The three methods are compared in terms of goodness of fit and prediction accuracy. Deviance and log-likelihood are used as the metrics for the goodness of fit, and root mean square error and normalized root mean square error are used as the metrics for prediction accuracy. The accurate estimation of recovery parameters from disasters improves recovery management, such as the optimization of resource allocation. We illustrate this work using a case study of Shelby County in Tennessee.

Hiba Baroud
Vanderbilt University
United States

Jinzhu Yu
Vanderbilt University
United States

 

Powered by OpenConf®
Copyright ©2002-2016 Zakon Group LLC