The evaluation of uplift models differs much from other predictive analytics approaches. This node evaluates the quality of an Uplift Model by analyzing the real uplift in given bins compared to the predicted uplift. The best way to do so, is to convert and append the predicted uplift into a binned column. In most cases an Uplift Tree will only have a certain amount of possible outcomes (e.g. 25 different prediction values). These 25 values are considered as 25 bins. The Uplift Tree Evaluator now calculates the real uplift within the bins and compares it to the predicted uplift of the model. Similar uplift values will result in a good performance of the uplift model.
- Type: TableInputInput data with contact group, target group, binned real uplift and predicted uplift columns.