The treatment effect on a set of observations is the {{ isCausalClassification() ? 'preferred outcome class frequency' : 'average outcome' }} for its treated observations minus the {{ isCausalClassification() ? 'preferred outcome class frequency' : 'average outcome' }} for its control observations.
The cumulative uplift is the treatment effect on the subset defined by the highest ranking elements according to the predictions of treatment effect, up to a specified ratio of the test set. The net uplift is the observed cumulative uplift minus the expected cumulative uplift obtained by a random treatment assignment.
For different fractions of the test observations, the values of the cumulative uplift curve are computed by dividing the average uplift by the number of test observations. These values are then normalised by the total number of test observations.
The values of the Qini curve are computed by dividing the average uplift by the number of treated test observations. These values are then normalised by the total number of treated test observations.
The Area Under the Uplift Curve is the area between the cumulative uplift curve and the dashed line corresponding to a random treatment assignment.
The Qini score is the area between the Qini curve and the dashed line corresponding to a random treatment assignment.
The Net uplift at {{ mlTaskDesign.modeling.metrics.netUpliftPoint * 100 | nicePrecision: 4 }}% is the difference between the value on the cumulative uplift curve and the one on the random treatment assignment dashed line, for x = {{ mlTaskDesign.modeling.metrics.netUpliftPoint * 100 | nicePrecision: 4 }}%.