Shivaram Subramanian

Counterfactual Optimal Action Trees (COAT): Interpretable Prescriptive Policies from Observational Data

Youssef Drissi, Markus Ettl, Shivaram Subramanian, Wei Sun, Zack Xue

We introduce COAT (Counterfactual Optimal Action Tree), a framework for learning interpretable prescriptive policies from observational data. COAT combines counterfactual outcome estimation with large-scale mixed-integer optimization, using column generation to translate causal p…Read more