Fetching the paper…

Reading the bibliography…

2017

Consistent feature attribution for tree ensembles

Lundberg, Scott M., Lee, Su-In

Original
PDF
Post on X

Understand

Note that a newer expanded version of this paper is now available at: arXiv:1802.03888 It is critical in many applications to understand what features are important for a model, and why individual predictions were made.

For tree ensemble methods these questions are usually answered by attributing importance values to input features, either globally or for a single prediction.
Here we show that current feature attribution methods are inconsistent, which means changing the model to rely more on a given feature can actually decrease the importance assigned to that feature.
To address this problem we develop fast exact solutions for SHAP (SHapley Additive exPlanation) values, which were recently shown to be the unique additive feature attribution method based on conditional expectations that is both consistent and locally accurate.

Consistent feature attribution for tree ensembles · Around

Built on

Classification and regression trees
Breiman, Leo, Friedman, Jerome, Stone, Charles J, and Olshen, Richard A · 1984
Earlier work this paper cites.
The consortium to establish a registry for alzheimer’s disease (cerad) part ii. standardization of the neuropathologic assessment of alzheimer’s disease
Mirra, Suzanne S, Heyman, A, McKeel, D, Sumi, SM, Crain, Barbara J, Brownlee, LM, Vogel, FS, Hughes, JP, Van Belle, G, Berg, L, et al · 1991
Earlier work this paper cites.
The elements of statistical learning , volume 1
Friedman, Jerome, Hastie, Trevor, and Tibshirani, Robert · 2001
Earlier work this paper cites.
Generalized boosted regression models. documentation on the r package ‘gbm’, version 1.6–3, 2010
Ridgeway, Greg · 2010
Earlier work this paper cites.

Similar

Scikit-learn: Machine learning in python
Pedregosa, Fabian, Varoquaux, Gaël, Gramfort, Alexandre, Michel, Vincent, Thirion, Bertrand, Grisel, Olivier, Blondel, Mathieu, Prettenhofer, Peter, Weiss, Ron, Dubourg, Vincent, et al · 2011
Cited alongside, same era.
Efficient dimensionality reduction for high-dimensional network estimation
Celik, Safiye, Logsdon, Benjamin, and Lee, Su-In · 2014
Cited alongside, same era.
Xgboost: A scalable tree boosting system
Chen, Tianqi and Guestrin, Carlos · 2016
Cited alongside, same era.

Then

An unexpected unity among methods for interpreting model predictions
Original
Lundberg, Scott and Lee, Su-In · 2016
Later among the works it cites.
A unified approach to interpreting model predictions
Original
Lundberg, Scott and Lee, Su-In · 2017
Closest in time.
Interpreting random forests
Saabas, Ando · 2017
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…