Aggregation Methods for Large-Scale Location Problems

配送领域和城市服务领域的许多问题需要对设施位置的最佳选择做出决策，包括仓储、服务中心和零售决策。计算机模型可以用来帮助做出这些决定。这样的计算机模型通常需要关于客户(称为需求点)在哪里、有多少客户以及他们的需求是什么的信息。在许多计算机模型应用中，请求点可以数以百万计。即使所有的需求点数据都可用，通常也不可能将其全部包含在模型中。相反，请求点通常是聚合的；例如，可以假设一个邮政编码区域中的所有请求点都位于邮政编码区域的中心。这种聚合减小了模型的大小，但也会产生建模误差。实践中通常使用的聚合方案通常是特别的，很少或根本不利用问题结构，并且通常不关心它们的聚合错误。该建议解决了进行请求点聚合的方法，以便将误差保持在可管理的限度内。对于各类选址问题，在前人工作的基础上提出了集结方法的发展。将研究影响误差的模型参数，并对聚合方法进行计算机测试。研究应该有助于积累知识体系，最终将导致更好地理解和解决需求点聚合问题。位置分析员将能够更好地平衡计算机模型所需解决方案的质量与聚合带来的误差。当设施选址的决策是基于计算机建模时，研究结果将有助于更好地做出此类决策。研究结果对城市和区域规划者、运输/物流专家和地理学家是有用的，他们有时都会参与选择大规模选址问题的聚集方法。

Numerous problems in the area of distribution and in urban services require decisions about best choices of facility locations, including warehousing, service centers, and retailing decisions. Computer models can be used to help with these decisions. Such computer models usually require information on where the customers (referred to as demand points) are, how many of them there are, and what their demands are. In many computer model applications, the demand points can number in the millions. Even if all the demand point data is available, it is usually infeasible to include it all in the models. Instead, demand points are usually aggregated; for example, all the demand points in one postal code area may be assumed to be at the center of the postal code area. This aggregation reduces the size of the model, but also creates modeling error. Aggregation schemes typically used in practice are usually ad-hoc, make little or no use of the problem structure, and generally show no concern for them aggregation error. This proposal addresses means of doing demand point aggregation so as to keep the error to manageable limits. For various classes of location problems, the development of aggregation methods that build on previous work is proposed. Model parameters that affect the error will be studied, and computer testing of the aggregation methods will be performed. The research should contribute to a cumulative body of knowledge that will eventually result in demand point aggregation problems being better understood and solved. Location analysts will be better able to balance the quality of the solutions needed from computer models with the error introduced by the aggregation. The results of the research should be helpful in allowing better decisions about facilitysiting when such decisions are based on computer modeling. The results should be useful to urban and regional planners, transportation/logistics specialists, and geographers, all of whom at times are involved in choosing aggregation methods for large-scale location problems.