Parameter-Free Spatial Data Mining Using MDL

Free registration required

Executive Summary

Consider spatial data consisting of a set of binary features taking values over a collection of spatial extents (grid cells). It proposed a method that simultaneously funds spatial correlation and feature co-occurrence patterns, without any parameters. In particular, it employs the Minimum Description Length (MDL) principle coupled with a natural way of compressing regions. This defines what "Good" means: a feature co-occurrence pattern is good, if it helps better compress the set of locations for these features. Conversely, a spatial correlation is good; if it helps better compress the set of features in the corresponding region. Its approach is scalable for large datasets (both number of locations and of features).

  • Format: PDF
  • Size: 269.6 KB