Clustan

from Wikipedia, the free encyclopedia

Clustan is one of the oldest integrated statistics programs for cluster analysis and data mining . The beginnings go back to the Fortran programs from 1968. David Wishart is also the developer and head of Clustan Ltd from Edinburgh . In the 1990s, he adapted the software to the (Windows) PC and provided it with interactive graphics.

Topics of the clustan software

Hierarchical cluster methods

  • Cluster: hierarchical cluster analysis on data matrices
  • Hierarchy: hierarchical cluster analyzes on similarity matrices
  • Centroid: Focus formation in similarity matrices
  • Density: hierarchical density estimator method
  • Divide: hierarchical distribution clustering on binary variables

Non-hierarchical cluster methods

  • Classify: identification of new cases by traversal of cluster trees
  • Normix: Maximum likelihood estimation of the measures of multivariate normal distributions with significance tests
  • Invariant: iterative optimization of Wilks' Lambda or Hotelling's Trace
  • Mode: Finding “natural” clusters by estimating disjoint density areas
  • Relocate: iterative assignment to clusters ( k-means algorithm )
  • Kdend: Search for Bk-overlapping-clusters (Jardine-Sibson cluster method , see Voronoi interpolation )
  • Dndrite: Division of minimally reaching trees in order to minimize the sum of the squares of errors ( regression analysis )
  • Euclid: fuzzy clustering to minimize sums of squares

Other statistical and graphic auxiliary routines

  • Read similarity matrix: Reading of similarity matrices
  • Calculate similarity matrix: Calculate a similarity matrix with a large number of different coefficients
  • Print results: Print the cluster analysis results
  • Scatter: Draw scatter and cluster diagrams ( plots )
  • Plink: drawing hierarchical cluster trees
  • Rules: significance tests for the best division
  • Compare: comparison of hierarchical classifications

Graphic display software

  • ClustanGraphics: interactive software to view the results graphically

See also

Web links