Interactive Visual Methods for Partitioning Multidimensional Spatial Data

University Of Maryland Baltimore County, Baltimore MD

Investigators

Abstract

The goal of this research project is to develop innovative tools for interactive visual exploration of spatial multivariate data, using methods from artificial intelligence and information visualization. The tools is motivated by the problem of school redistricting, and is conducted in collaboration with staff from the Howard County (Maryland) Public School System. The approach developed in this project produces multiple similar solutions with respect to optimization criteria, but trying to ensure that the solutions are qualitatively different. The interactive environment allows the user to browse through "nearby" solutions, investigate minor perturbations of each solution, while reducing the number of critically different solutions. This interactive visual method is expected to be very effective in the school redistricting domain, resulting in a substantial reduction in the time to develop new redistricting plans, and a corresponding increase in the number of plans that can effectively be generated and compared. This will result improved school redistricting process, where the visual tools will improve the ability of the school system to explain, justify, and disseminate proposed plans and the associated quantitative evaluation. The methods developed in this project will also be applicable in almost any other domain where multidimensional spatial data need to be partitioned according to some criteria. The project's Web site (http://maple.cs.umbc.edu/redistricting/) will be used to disseminate the results of this research.