2024 Undergraduate Research Showcase

Discovering Correlative Relations in Spatial Phenomena

Document Type

Student Presentation

Presentation Date

4-19-2024

Faculty Sponsor

Dr. Jared Talley

Abstract

Correlation does not imply causation, but it can provide useful insights when phenomena are very complex. We are living in a time of abundant geospatial information. We present a novel approach for discovering correlative relations from geospatial data.

Geospatial data consists of areas and points on a map. We consider the ratio between the quantity of points in an area, and the size of an area; The density. Dense areas are "promising leads”, and sparse areas are "dead ends". Boolean operations are used to discover relations between areas. We consider OR, AND, NOT, and XOR. These operations are applied to every combination of input, yielding new areas.

The algorithm repeats the process of evaluating areas and generating new areas. Step by step, it builds areas with better densities. Borrowing from Darwin, each step produces a generation of areas. The density function tells us the fitness of an area. After many steps, we’re left with the final set of areas. Since each area remembers its parents, the algorithm can trace the lineage of every area. If many of these areas share a common ancestor, then a quality of that ancestor makes it more "fit".

This document is currently not available here.

Share

COinS