Predicting Hurricane Evacuation Decisions with Interpretable Machine Learning Methods

Facing the escalating effects of climate change,it is critical to improve the prediction and understanding of the hurricane evacuation decisions made by households in order to enhance emergency management.Current studies in this area often have relied on psychology-driven linear models,which frequently exhibited limitations in practice.The present study proposed a novel interpretable machine learning approach to predict household-level evacuation decisions by leveraging easily accessible demographic and resource-related predictors,compared to existing models that mainly rely on psychological factors.An enhanced logistic regression model(that is,an interpretable machine learning approach) was developed for accurate predictions by automatically accounting for nonlinearities and interactions(that is,univariate and bivariate threshold effects).Specifically,nonlinearity and interaction detection were enabled by low-depth decision trees,which offer transparent model structure and robustness.A survey dataset collected in the aftermath of Hurricanes Katrina and Rita,two of the most intense tropical storms of the last two decades,was employed to test the new methodology.The findings show that,when predicting the households’ evacuation decisions,the enhanced logistic regression model outperformed previous linear models in terms of both model fit and predictive capability.This outcome suggests that our proposed methodology could provide a new tool and framework for emergency management authorities to improve the prediction of evacuation traffic demands in a timely and accurate manner.

Design of an Evacuation Demand Forecasting Module for Hurricane Planning Applications

This paper discusses the development and implementation of an evacuation demand forecasting module that was incorporated into a comprehensive decision support system for the planning and management of contraflow operations in the Gulf of Mexico. Contraflow implies the reversing of one direction of a highway in order to permit a substantially increased travel demand exiting away from an area impacted by a natural disaster or any other type of catastrophic event. Correctly estimating the evacuation demand originated from such a catastrophic event is critical to a successful contraflow implementation. One problem faced by transportation officials is the arranging of the different stages of this complex traffic procedure. Both the prompt deployment of resources and personnel as well as the duration of the actual contraflow affect the overall effectiveness, safety and cost of the evacuation event. During this project, researchers from the University of Alabama developed an integral decision support system for contraflow evacuation planning to assist the Alabama Department of Transportation Maintenance Bureau in the evaluation and planning of contraflow operations oriented to mitigate the evacuation burdens of a hurricane event. This paper focuses on the design of the demand forecasting module of such a decision support system.

An integer multi-objective optimization model and an enhanced non-dominated sorting genetic algorithm for contraflow scheduling problem

To determine the onset and duration of contraflow evacuation, a multi-objective optimization(MOO) model is proposed to explicitly consider both the total system evacuation time and the operation cost. A solution algorithm that enhances the popular evolutionary algorithm NSGA-II is proposed to solve the model. The algorithm incorporates preliminary results as prior information and includes a meta-model as an alternative to evaluation by simulation. Numerical analysis of a case study suggests that the proposed formulation and solution algorithm are valid, and the enhanced NSGA-II outperforms the original algorithm in both convergence to the true Pareto-optimal set and solution diversity.