Abstract:
|
We seek to address the growing need for data driven pricing policies in the business context. Often times policy makers are left with gut-feel pricing decisions which may not be optimal. When policy makers seek policy prescriptions from opaque, black-box machine learning models they often run into an insuperable dilemma: how to convince a broader stakeholder buy-in? With interpretable prescriptive models, policy makers can understand how the decisions are being made, agree or disagree with them, explain them to their colleagues, and bring transparency to the decision-making logic. In this way we understand that interpretability is not only a “nice-to-have” but a necessary and vital component in applications of machine learning in the business decision making context. The research currently being conducted is focused on optimal pricing for revenue maximization such as seen in Biggs et al. (2021) \cite{biggs2021model}, while reviewing a select set of algorithms, including, Personalization Trees (Kallus 2017) \cite{kallus2017recursive}, Causal Forests (Wager and Athey 2018) \cite{wager2018estimation} and Optimal Prescription Trees (Bertsimas et al., 2019) \cite{bertsimas2019optimal},
|