Semester

Spring

Date of Graduation

2012

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Arun A. Ross.

Abstract

Pareto Optimization is a method of finding the space of best solutions possible given multiple class dimensions. In practice, however, it is difficult for users to visualize how solutions are spaced across this multi-dimensional frontier. Existing algorithms like NSGA-II are capable of quickly finding the Pareto frontier. However, users cannot use these algorithms to understand how changes to a solution effect the qualities of that solution.;In response to these issues, this paper presents "HOW". Not only is "HOW" capable of finding the Pareto frontier across multiple dimensions, but HOW can also explain how each point was derived, and compare points along the frontier to their neighbors. HOW generates rules using a stochastic learner, then combines those rules into a subsumption network via formal concept analysis. The network is augmented with statistics on the training data selected by sub-networks. Users can navigate their decision space by walking that network, selecting regions containing properties they want to avoid or encourage.

Share

COinS