Databases Reference
In-Depth Information
distribution, it is a strong indicator that in subsequent exploration and algorithm
application, we should attempt to understand why. (Note: Smoothing of a
distribution can be decreased and returned to its unsmoothed level by repeatedly
clicking I the “
” button.
The scatter plot
A third available viewer in exploring a dataset is the scatter plot - useful for
evaluating the nature of relationships between attributes. The scatter plot is
probably a familiar plot to you as it represents observations as points on X-Y
(and Z) axes.
To open a scatter plot of the Iris dataset, drag the dataset up to the icon
representing the display currently being used for the correlation matrix. As
you drag the dataset over the display icon, a dashed rectangle is drawn
showing you where the new plot will be created. As you drag to the left, the
rectangle will be on the left pushing the current correlation matrix to the
right side of the display (Figure 2.11). As you drag to the right, you will
first see the rectangle fill the entire display, indicating that the new plot will
replace the correlation matrix. Continuing to the right, the rectangle moves
to the right side of the display pushing the correlation matrix to the left.
Drop the dataset at this location.
Select “Scatter Plot” from the context menu. The plot is opened and the
correlation matrix is pushed to the left side of the display (Figure 2.12).
By default the scatter plot shows the first two attributes in the dataset on the X
and Yaxes respectively. Notice that there is no scale on either of the axes. This is
intentional. In VisMiner, scatter plots are intended to represent relationships
between attributes, not to be used for point reading. The only indicator of scale
Figure 2.11
Viewer Location
