A modeling strategy for large-scale optimization based on analysis and visualization principles

Abstract

A modeling strategy for the validation and analysis of large-scale optimization models is defined and demonstrated. The strategy is based on nine principles of analysis and eight principles of visualization that are applied in a user controlled hierarchical structure which is customized to a particular optimization problem. For each model a set of analytic tools, such as spreadsheets and graphs, is structured to validate and verify data and analyze the model and its results. These tools can be quickly recreated with data from subsequent runs of the model and sensitivity analysis conducted and comparisons made. As a demonstration the strategy is applied to PHOENIX, a large-scale U.S. Army helicopter force planning model. the strategy incorporates available technology using commercially prepared software and a computer workstation. the applications of techniques such as hypertext, data access and backward compatibility enhance the ease of use and effectiveness of this approach.