Databases Reference
In-Depth Information
Figure 5.4
Classification Surface - Iris data
of the algorithms. The “Classification Model” viewer was designed to support
this task.
Drag the SVM model for the Iris dataset up to an available display, release
and select “Classification Surface”.
The visualization that opens (Figure 5.4) shows an XYZ surface plot of the
likelihoods of the selected output attribute value given the input attributes.
The likelihoods, as plotted on the Yaxis, range in value between zero and one.
As with other 3-D viewers in VisMiner, the plot may be rotated to gain a better
perspective of the actual surface shape. Scales on the axes are omitted in order
to focus the user's attention on the surface shape rather than point reading.
In the plot you can see that the likelihood of being variety Setosa is greatest
for low values of PetalLength and PetalWidth. As both the length and width
increase the likelihood drops.
Notice that the point of greatest likelihood is only about two-thirds of the way
up the Y axis - close to about 0.7 likelihood. You might be wondering if there
aren't combinations of PetalWidth and PetalLength that have greater like-
lihoods. The answer can be found in the option panel to the right (Figure 5.5).
There are four sliders - one for each of the input attributes that were used to
build the model. The top two, for PetalLength and PetalWidth are disabled,
because they are currently selected on the X and Z axes respectively. The slider
for SepalLength is set at 6.1 and the slider for SepalWidth is set at 3.2. The
surface plot that you see represents possible combinations of PetalLength and
PetalWidth when SepalLength is fixed at 6.1 and SepalWidth is fixed at 3.2.
Since SepalLength and SepalWidth were used to build the model, they also
contribute to the output likelihoods.
Drag the slider for SepalLength to the left.
