The Lift Chart of GeneXproTools shows both the Lift Curve and
Cumulative Lift Curve on the same graph. These curves are also shown in relation to a
Random Model and an Ideal Model.
The Random line in the Lift Chart represents the average
response rate. And the Ideal line represents a perfect model
that is never wrong and therefore could select all the estimated
positive responses. The point where the Lift Curve crosses the Random line corresponds approximately to the percentage of the population beyond which the benefits from using the model are lost.
Other useful visual clues from the Lift Chart include the Area Between both
Lift Curves (represented by
ABC in the Lift Chart). Theoretically, the greater ABC the better the model. The individual areas under each of the
Lift Curves are also computed and shown both on the Lift Chart and in the companion
Lift Statistics Report.
The Lift Ranking Quality (LRQ) is yet another useful indicator of the accuracy of a model. It corresponds to the ABC area normalized against the area under the Ideal
line. Negative values both for the ABC and LRQ are indicative of an inverted model.