Optimizing Warehouse Distribution by Random Sampling Using Census Data
Abstract
Businesses and governments need optimal locations for new buildings and services. The goal of this study was to model warehouse (any building or service local) distribution and use that model to study the relationship between warehouse distribution and population distribution. Warehouse distribution was modeled using Census tract and block data. To calculate the best location for a set of warehouses in a given geographic area, a random sampling optimization algorithm was run to improve the average distance a person would have to travel to their nearest warehouse. The qualities of many states and countries' warehouse distribution were studied by running the optimization algorithm with varied parameters. The U.S. national warehouse distribution was analyzed by performing the algorithm for varied sets of 1-2000 warehouses and plotting the relationship between warehouse count and the average distance a consumer would have to travel. The U.S. distribution was compared with San Francisco, Tennessee, Australia, and South Africa distributions to study the relationship between a population distribution's shape and its warehouse count vs. minimum average distance distribution. The wider a population is dispersed the closer the exponent of the warehouse count vs. minimum average distance trendline is to 2, while the exponent is closer to 1 for more one-dimensional, linear population distributions. Using this algorithm to understand the relationship between optimal business and service locations and population distribution could lead to better urban planning and less service deserts.
Optimizing Warehouse Distribution by Random Sampling Using Census Data
HSS 209
Businesses and governments need optimal locations for new buildings and services. The goal of this study was to model warehouse (any building or service local) distribution and use that model to study the relationship between warehouse distribution and population distribution. Warehouse distribution was modeled using Census tract and block data. To calculate the best location for a set of warehouses in a given geographic area, a random sampling optimization algorithm was run to improve the average distance a person would have to travel to their nearest warehouse. The qualities of many states and countries' warehouse distribution were studied by running the optimization algorithm with varied parameters. The U.S. national warehouse distribution was analyzed by performing the algorithm for varied sets of 1-2000 warehouses and plotting the relationship between warehouse count and the average distance a consumer would have to travel. The U.S. distribution was compared with San Francisco, Tennessee, Australia, and South Africa distributions to study the relationship between a population distribution's shape and its warehouse count vs. minimum average distance distribution. The wider a population is dispersed the closer the exponent of the warehouse count vs. minimum average distance trendline is to 2, while the exponent is closer to 1 for more one-dimensional, linear population distributions. Using this algorithm to understand the relationship between optimal business and service locations and population distribution could lead to better urban planning and less service deserts.
