Early Insights into Location-Allocation Decision-Making using Ensemble Learning

Abstract

The COVID-19 pandemic underscored the significance of geography in health emergencies, spotlighting the need for optimized spatial decision-making. This paper introduces a novel, data-driven approach to spatial decision-making, leveraging gradient boosting to derive datainformed weights for a Weighted Linear Combination (WLC). The goal is to pinpoint optimal locations for vaccination centres in Flanders, Belgium. Drawing from prior work, we present a foundation for the required number of centres, and then focus on determining the most suitable locations within Flanders. Utilizing a dataset of 91 centres, our ensemble learning technique dynamically determines criteria weights. Criteria that are sociodemographic or mobility oriented are considered. Our methodology, termed Ensemble Analysis for Criteria Trade-offs (ENACT), offers a comprehensive framework, targeting dynamic location-allocation scenarios. Using derived weights, we identify regions with the highest suitability scores for vaccination center placement. The high model performance metrics underline its reliability, with caution on potential overfitting. The study comes with a roadmap for enhancing the methodology's comprehensiveness in future research, suggesting the integration of more criteria and GIS optimization techniques for actionable health infrastructure planning.