QPRC 2016

A Layered Pareto Front Approach to Search for the Top N Subpopulations in a Stockpile

Sarah Burke, Arizona State University
Christine Anderson-Cook, Los Alamos National Laboratory
Lu Lu, University of South Florida
Connie Borror, Arizona State University West
Douglas Montgomery, Arizona State University
                                                              
Abstract


In a decision-making process, relying on only one objective or criterion can lead to oversimplified sub-optimal decisions. Incorporating multiple, and likely competing, objectives is critical to balance the tradeoffs between the criteria. There are many situations where we may be interested in several optimal solutions, not just one. These may fall into one of two scenarios: 1) decision-makers need to identify the best N solutions to accomplish a goal or specific task, or 2) a decision is evaluated based on several primary objectives we can formally quantify along with secondary, qualitative priorities. In this second scenario, we can identify several contending solutions (and eliminate non-contenders) using the primary, quantitative objectives. We then use the secondary, qualitative objectives to make the final decision. In this talk, we present a two-phase layered Pareto Front method to search for the top N solutions of a given problem. In the first objective phase, layered Pareto Fronts are identified, eliminating the dominated choices from further consideration. In the second phase, the remaining solutions from phase 1 are ranked from 1 to N for different priorities of the criteria to identify the top N solutions under different scenarios. The method and accompanying JMP Add-In are demonstrated to identify the top N critical subpopulations in a stockpile. This program also provides several graphical tools to assist the decision-maker in making their final decision.