Spatial autocorrelation informed approaches to solving location-allocation problems

Keywords: location-allocation, p-median, spatial autocorrelation, spatial median, spatial optimization
Session Type: Virtual Paper Abstract
Day: Friday
Session Start / End Time: 2/25/2022 03:40 PM (Eastern Time (US & Canada)) - 2/25/2022 05:00 PM (Eastern Time (US & Canada))
Room: Virtual 59

Abstract

Surveying programs of study at institutions of higher learning throughout the world reveals that one natural disciplinary coupling is statistics and operations research, although these two specific disciplines currently lack an active synergistic research interface. Similarly, the development of spatial statistics and spatial optimization has occurred in parallel and nearly in isolation. This paper seeks to alter this situation by initiating transformative work at the interface of these two subdisciplines, encouraging considerably more future interaction between them. It outlines three ways spatial statistics can contribute to spatial optimization by exploiting spatial autocorrelation in georeferenced data: missing attribute value imputation (analogous to kriging); identifying colocations of local spatial autocorrelation hotspots and spatial medians; and, geographic tessellation stratified random sampling inputs to spatial optimization heuristics that successfully guide them to optimal location solutions. One contention emphasized throughout this paper is that this spatial statistics/optimization interface furnishes another vehicle for delivering spatial statistical benefits to society, which, in turn, benefits spatial statistics by providing better integration of it into novel interdisciplinary contexts.