Technical Report: Partial Dependence through Stratification

Partial dependence curves (FPD) introduced by Friedman, are an important model interpretation tool, but are often not accessible to business analysts and scientists who typically lack the skills to choose, tune, and assess machine learning models. It is also common for the same partial dependence algorithm on the same data to give meaningfully different curves for different models, which calls into question their precision. Expertise is required to distinguish between model artifacts and true relationships in the data. In this paper, we contribute methods for computing partial dependence curves, for both numerical (StratPD) and categorical explanatory variables (CatStratPD), that work directly from training data rather than predictions of a model. Our methods provide a direct estimate of partial dependence, and rely on approximating the partial derivative of an unknown regression function without first fitting a model and then approximating its partial derivative. We investigate settings where contemporary partial dependence methods---including FPD, ALE, and SHAP methods---give biased results. Furthermore, we demonstrate that our approach works correctly on synthetic and plausibly on real data sets. Our goal is not to argue that model-based techniques are not useful. Rather, we hope to open a new line of inquiry into nonparametric partial dependence.

[1]  W. Cleveland,et al.  Locally Weighted Regression: An Approach to Regression Analysis by Local Fitting , 1988 .

[2]  Robert Chen,et al.  Machine Learning Model Interpretability for Precision Medicine , 2016, 1610.09045.

[3]  Achim Zeileis,et al.  BMC Bioinformatics BioMed Central Methodology article Conditional variable importance for random forests , 2008 .

[4]  Carlos Guestrin,et al.  "Why Should I Trust You?": Explaining the Predictions of Any Classifier , 2016, ArXiv.

[5]  Leo Breiman,et al.  Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author) , 2001 .

[6]  Scott Lundberg,et al.  An unexpected unity among methods for interpreting model predictions , 2016, ArXiv.

[7]  J. Friedman Greedy function approximation: A gradient boosting machine. , 2001 .

[8]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[9]  W. Cleveland LOWESS: A Program for Smoothing Scatterplots by Robust Locally Weighted Regression , 1981 .

[10]  Dominik Janzing,et al.  Feature relevance quantification in explainable AI: A causality problem , 2019, AISTATS.

[11]  D. Donoho 50 Years of Data Science , 2017 .

[12]  Daniel W. Apley,et al.  Visualizing the effects of predictor variables in black box supervised learning models , 2016, Journal of the Royal Statistical Society: Series B (Statistical Methodology).

[13]  Galit Shmueli,et al.  To Explain or To Predict? , 2010 .

[14]  Been Kim,et al.  Towards A Rigorous Science of Interpretable Machine Learning , 2017, 1702.08608.

[15]  Emil Pitkin,et al.  Peeking Inside the Black Box: Visualizing Statistical Learning With Plots of Individual Conditional Expectation , 2013, 1309.6392.

[16]  W. Cleveland Robust Locally Weighted Regression and Smoothing Scatterplots , 1979 .

[17]  Mukund Sundararajan,et al.  The many Shapley values for model explanation , 2019, ICML.

[18]  Paulo J. G. Lisboa,et al.  Making machine learning models interpretable , 2012, ESANN.

[19]  Giles Hooker,et al.  Please Stop Permuting Features: An Explanation and Alternatives , 2019, ArXiv.