The following figure shows the Text View of the Clustering Visualizer:
Each row in the table describes one cluster by showing how the field values are distributed for the records in this cluster.
The background color of a row represents a particular information of the corresponding cluster, for example, the size of a cluster. You can also specify the background color to represent the average values of a specific field, or aggregations of all record values in the class of a specific field. You can specify the background color and the sorting criteria on the Color Coding page of the Properties notebook.
In the graphic above, the clusters are sorted by size. The largest cluster is displayed on top of the window. The legend shows that the cluster colors are determined by their average salaries.
- Categorical
- A categorical string is displayed. For example, Marital_Status is
predominantly m.
The state is determined by the most frequent field value in the cluster. For example, looking at cluster 5 in the figure above, the contribution of the field MARITAL_STATUS to the cluster description is MARITAL_STATUS is predominantly s.
- Continuous numerical
- A value is mapped
to low, medium, or high, or to the border numbers
of the field specification.
The state is determined by the interval with the highest frequencies in that cluster related to the frequencies in the whole population. If statistical information is available, the labels low, medium, or high are used to determine the state. For example, in the figure above in cluster 1, the state for the field Salary is determined as high.
- Discrete numerical
- The numeric value is displayed as a string. For example, Year_1st_Policy is predominantly 1991.
For continuous numerical fields, the fields are described by the intervals having the highest frequency. If statistical information is available, the labels low, medium, or high are used to determine the state.