Geo Mining in ELKI

Spatial data mining in ELKI is supported at various places.

Specialized Geo data mining algorithms

• Outlier detection:
  • CTLuGLSBackwardSearchAlgorithm
  • CTLuMeanMultipleAttributes
  • CTLuMedianAlgorithm
  • CTLuMedianMultipleAttributes
  • CTLuMoranScatterplotOutlier
  • CTLuRandomWalkEC
  • CTLuScatterplotOutlier
  • CTLuZTestOutlier
  • SLOM
  • SOF
  • TrimmedMeanApproach

Additionally, you can use existing data mining Algorithms via specialized DistanceFunctions.

Geo distance functions

Defined for latitude-longitude pairs:

• DimensionSelectingLatLngDistanceFunction
• LatLngDistanceFunction
• LngLatDistanceFunction

See DistanceFunctions for other distance functions.

Geo data types

• Regular number vectors (e.g. 2-d with longitude and latitude meaning)
• Multi-Polygons

See DataTypes for a complete overview.

Index support

Metrical IndexStructures have support for geo data by using geographic distance functions.

Support in R-Trees is not yet complete, as we do not yet have an implementation of a valid min-dist for spatial rectangles on the earth sphere.

Output support

ELKI has some specialized output support for geo data:

• KMLOutputHandler for producing KML output files
• JSONResultHandler for using a web browser interface