a month ago
Hello, I am interested in being able to use the techniques of uplift (true response, net modeling, etc.) model, but I do not find in their knowledge base, forums or samples.
I understand that it is key for these models to use control groups in the last stage of design but I do not know how to apply them.
Do you have an example of use or any documentation where it is detailed with Rapidminer?
Thanks in advance.
2 weeks ago
hello @RAC - sorry no one has chimed in here. Have you looked at the Lift Chart Template?
2 weeks ago
Thanks sgenzer I know it, but this only works for part of an uplift model. It would need to be integrated with a method for comparing the outcome against a control group (Graph1). The difference between the two would be the gain and would result in a real uplift different from the one not used by the control group.
For example, for a customer segment they can give many positive responses as a result of a campaign, but if compared to a control group that has not received such a campaign and represents the same segments, it may be possible to discover that without such a campaign the same result (false positives) or even worse (false negatives). These should be discarded from the entire action (Graph2).The goal is to discard these from the entire action.