Cluster Analysis

Cartographica can perform basic and advanced cluster analysis based on a number of parameters. The overall algorithm used is DBSCAN, which is a progressive algorithm that is efficient and can find cluster patterns that elude other algorithms with a minimal amount of pre-configuration. This makes the algorithm particularly good for experimental or preliminary analysis, while still remaining effective for detailed analysis.

Cluster Analysis is for dealing with Point features, and cannot be used with other geometry types.

Performing a Cluster Analysis
  1. Select the Layer that contains the points you want to cluster
  2. Choose  Tools > Find Clusters…
    The  Find Clusters sheet appears.

    Find Clusters sheet

    Shortly after the sheet comes up, the Average Distance (average nearest distance for points in this layer) and NNI (the Nearest Neighbor Index–a value describing how likely it is that the data forms clusters; if the value is <1 it is likely clustered).

  3. Adjust the  Minimum Count to be the minimum number of points in a cluster
  4. There are 3 distance functions that can be used in determining whether a point can be clustered with other points, of which you must choose one:
    1. Fixed distance will be in the coordinate system of the Map (which is consistent with Cartographica analysis tools)
    2. Average Nearest Neighbor Distance is based on the average nearest neighbor distance, and the value with the stepper indicates the number of standard deviations to include
    3. Expected Mean Distance is based on a confidence interval based on the the expected density of the area
  5. Finally, when you are satisfied with the parameters, click  Find
    The result layer will be added to your map as a new Polygon layer with a name based on the name of the layer to be analyzed

The techniques, and especially the Expected Mean Distance confidence intervals are modeled on those used by CrimeStat.