15. WORKLOAD CHARACTERIZATION › 15.3 Standard Output › 15.3.6 Cluster Index Analysis Graph

15.3.6 Cluster Index Analysis Graph

The Cluster Index Analysis graph can produce two different
charts, depending on data volumes and the print line size
specified in the MICF options:

   o BLOCK chart
   o Horizontal bar (HBAR) chart

In either case, one entry is made for each cluster selected
and the clustering index value is displayed.

                                      Workload Characterization Analysis
                                            Cluster Index Analysis
                                          For: Monday, June 23, 2003

                            Analysis Performed Using Variables: JOBTCBTM JOBEDASD

Cluster Number                                                                                  Cluster Index
                                                                                                          Sum
          |
      1   |                                                                                              0.00
          |
     10   |*****************                                                                             0.34
          |
     11   |***************************************                                                       0.77
          |
     14   |************************************************************                                  1.19
          |
     15   |***********************************************                                               0.95
          |
     16   |****************************************                                                      0.80
          |
     17   |*************************************************************************************         1.69
          |
     23   |*************************************************************************************         1.70
          |
     24   |*************************************************                                             0.98
          |
          -----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+
              0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70

                                               Cluster Index

                                   Sparse Clusters Have Been Excluded

 Figure 15-4B.  Cluster Index Analysis graph

The clustering index value is the Root Mean Square (RMS) of
the distances of all feature values within the cluster.  As
stated previously, this is a relative measurement of cluster
performance and "goodness of fit" of the data, with respect
to the cluster centroid values.  A lower value represents a
better performing cluster, indicating more tightly formed
clusters and fewer outlying values.

This graph is a simple, yet effective way to quickly evaluate
the overall performance of the clustering operation, and can
be used to communicate this performance to non-technical
personnel.