3. PERFORMANCE ANALYSIS TOOLS › 3.3 Data Clustering Analysis › 3.3.3 Standard Output

3.3.3 Standard Output

Data Clustering using Neugents technology produces reports
and simple graphs that describe the original data, show the
clusters that are identified from this data, and analyze the
degree of fit with which these clusters characterize the
data.

The 5 reports and 2 graphs are:

o  Clustering Input Descriptive Statistics.  This report
   details the characteristics of the variables that are
   specified as criteria for the clustering analysis.

o  Cluster Performance Summary.  This report consists of
   three sections:  Clustering Execution Options, a summary
   of the options selected by the analyst for the study;
   Cluster Feature Contents, a list of each of elements or
   features used for clustering, and the boundaries selected
   for determining normalcy of the data values; and finally,
   the Cluster Population Summary, which lists each cluster,
   its size, its percent of population, and an index used to
   indicate relative performance of the cluster.

o  Cluster Population Summary.  This report provides the
   resource consumption characteristics of each of the
   clusters.  For each cluster, its size and population are
   summarized and all clustering and reporting elements are
   listed, along with their respective usage values.  This
   report permits you identify resource consumption patterns
   that may represent performance or capacity problems.

o  Data Value Exception Detail.  This report provides a
   detail listing of any observation that has one or more
   clustering elements that are identified as outliers.  The
   report is sequenced in the same order as is the original
   CA MICS files used as input and can be used to further
   investigate why certain cases have been identified as
   outliers and whether or not they actually represent
   abnormal situations.

o  Sparse Cluster Population.  This report provides a listing
   of all clusters that have been identified as 'sparse' and
   their population contents.  A sparse cluster is defined
   within this application as one having less than 0.5% of
   the overall population of the study.  For example, if a
   sample of 2,000 cases were selected for analysis, then a
   sparse cluster would contain 10 or fewer observations.

o  Cluster Index Analysis.  This graph depicts a measurement
   of relative performance for each cluster, by presenting an
   index calculated by Neugents technology in terms of the
   root mean square of the distances of the resource vectors
   or feature values from the cluster centroid.  The cluster
   index of each cluster is displayed in either a Block or a
   horizontal bar chart, depending on the number of clusters
   being presented.  This chart permits the analyst to
   quickly access the overall performance of the workload
   characterization effort.

o  Cluster Population Analysis.  This graph depicts the
   population of the top "n" clusters in a PIE chart, which
   can be used to understand pictorially how the study data
   actually populated the clusters.


The following sections describe the fields, data elements,
and format of each report.

 1 - Clustering Input Descriptive Statistics Report
 2 - Cluster Performance Summary Report
 3 - Cluster Population Summary Report
 4 - Data Value Exception Detail Report
 5 - Sparse Cluster Population Report
 6 - Cluster Index Analysis Graph
 7 - Cluster Population Analysis Graph