CounterfactualUnitSelector Example

[aborundiya]

I was planing to use the CounterfactualUnitSelector to get the right audience for the upcoming campaign. Am bit lost as the CounterfactualUnitSelector has its own learner , how does the meta learner fits in this scheme ?

[aborundiya]

@t-tte @paullo0106 any thoughts on this ?

[t-tte]

Hi @aborundiya - the learner in the CounterfactualUnitSelector class is used to estimate the relevant conditional probabilities that Li and Pearl use to derive the bounds for the payoff from targeting a specific unit. No meta-learners are used in the class.

[aborundiya]

thanks @t-tte for the clarifications and was wondering why CounterfactualUnitSelector is in optimize library. I assumed that we may use the meta-learner to get the sensitive audience and use these audience as input to CounterfactualUnitSelector to get more refine audiences. What are your opinion on this ?

[t-tte]

If you want to use the output from eg a meta-learner to conduct value optimisation, you may want to look into the CounterfactualValueEstimator class: https://github.com/uber/causalml/blob/master/causalml/optimize/value_optimization.py Here is an example notebook using the approach: https://github.com/uber/causalml/blob/master/examples/counterfactual_value_optimization.ipynb