Databases Reference
In-Depth Information
Figure 14-14
Analysis Services 2005 provides two ways to analyze the validity of the mod-
el. One of them is to show the validity graphically and another is to show the
validation through actual numbers. To see the validity of the model graphic-
ally, select the Lift Chart sub-view within the Mining Accuracy Chart view. You
will now see the graph with two lines as shown in
Figure 14-15
.
The X-axis
shows the percentage of data set used for prediction and the Y-axis shows
the percentage of prediction correctness. You can see a legend window
providing details on the two lines. If you created a perfect model that pre-
dicted all inputs correctly, the percentage correctness will always be 100%.
This is represented by a blue line in the graph (the 45 degree line in
Figure
14-15
). In the current model the predictable attribute value can only have one
of the two values: 0 or 1. Obviously you would want a model that predicts val-
ues very close to that of an ideal model. This graph gives you a visual way to
easily compare the prediction correctness of the model as compared to an
ideal model. The prediction correctness percentage of the model will be
shown in a red line. You can see from
Figure 14-15
that the prediction cor-
rectness of the model is not very close to that of an ideal model but is reason-
ably good since the prediction results are correct for 76% (population correct
% for 100% of overall population) of the overall data set.
Search WWH ::
Custom Search