A general framework for causal classification

In many applications, there is a need to predict the effect of an intervention on different individuals from data. For example, which customers are persuadable by product promotion? patients should be treated with certain type treatment? These typical causal questions involving or change in outcomes made intervention. The cannot answered traditional classification methods as they only use associations outcomes. personalised marketing, these often uplift modelling. objective modelling estimate effect, but its literature does not discuss when represents effect. Causal heterogeneity can solve problem, assumption unconfoundedness untestable So practitioners guidelines their applications using methods. this paper, we for set decision making problems, and differentiate it classification. We conditions resolved (and heterogeneity) also propose general framework classification, off-the-shelf supervised flexible implementations. Experiments have shown two instantiations work (causal modelling, competitive other