Discriminating Patterns for Empirical Discovery in Geospatial Data

Discriminating Patterns for Empirical Discovery in Geospatial Data

Presenter: Wei Ding Wei.Ding@umb.eduComputer Science DepartmentUniversity of Massachusetts Boston

The NSF Workshop on GeoSpatial and GeoTemporal informatics ∙ January 8-9, 2009

Wei Ding . UMass Boston . Discriminating Patterns 2

Geospatial Discriminating Patterns

Spatial Association Rules Spatial Co-Location Patterns Regional Association Patterns

Solved or Almost Solved Spatial Association Rules [Koperski & Han 1995]

Spatial Co-Location Patterns [Huang & Shekhar]

Regional Association Patterns [Ding & Eick]

[huang2004]

Wei Ding . UMass Boston . Discriminating Patterns 3

Failed & Missing Modeling causal relationship among

geo-variables Complex chain of mutual interactions Nonlinearity and spatial variability

Efficient pattern summarization Laziness: large amount of patterns “super-patterns”

Robust & practical knowledge discovery tools to build empirical models for domain experts

Wei Ding . UMass Boston . Discriminating Patterns 4

What is next?Provide the scientific community with an efficient tool for auto-analyzing the root causes behind observed patterns in geospatial data.