The Department of Economics Brown Bag Research Workshop series provides an opportunity for faculty to present early-stage research projects to collect feedback from colleagues.

Presenter: Nathan Palmer, Federal Reserve Board

Title: Explaining Machine Learning by Bootstrapping Partial Marginal Effects and Shapley Values

Abstract: Machine learning and artificial intelligence methods are often referred to as “black boxes” when compared to traditional regression-based approaches. Traditional linear models are interpreted through their marginal relationships as captured by their regression coefficients. We show that the same relationship can be described rigorously for any machine learning model by examining the slope of the partial dependence functions, which we call the partial marginal effect (PME). We prove analytically that the PME of an OLS model is equivalent to the OLS regression coefficients. Bootstrapping provides standard errors and confidence intervals around the point estimates of the PMEs. We apply the PME to a hedonic house pricing example and demonstrate that the PMEs of neural networks, support vector machines, random forests, and gradient boosting models reveal the non-linear relationships discovered by the machine learning models. Finally, we extend PME to a Shapley value decomposition and explore how it can be used to further explain model outputs.

Location: Seminar Room, BB 4.02.02

Contact hayri.arslan@utsa.edu to RSVP for the Economics Brown Bag Research Workshop series.