Collecting and Prioritizing Building Information via Online Microtasks for Tsunami Evacuation Shelter Selection

Abstract

Securing sufficient tsunami evacuation shelters is essential for reducing tsunami casualties, but designating new facilities is costly for local governments. An alternative is to use existing private buildings, though potential candidates reach tens of thousands. While crowdsourcing is promising, its practical quality remains unclear. We set two research questions and conducted crowdsourcing experiments: (1) How feasible is it to collect reliable building information through online micro-tasks in crowdsourcing? (2) Are there any batch strategies to choose candidate buildings in order to maximize evacuation population coverage while accounting for possible refusals under the uncertainty? To answer the questions, we evaluated the quality of crowd-collected building information. Then, we formulated the task as a Maximum Coverage Problem with domain-specific constraints and analyzed an approximation algorithm with theoretical guarantees. Results show that crowdsourcing yields usable data quality and that the proposed prioritization approach is effective, demonstrating the feasibility of crowdsourcing-based evacuation shelter planning.