Add Microsoft EBM (Explainable Boosting Machine)

[mratsim]

Microsoft pre-released the Explainable Boosting Machine 2 weeks ago:

https://github.com/microsoft/interpret

It has very promising performance profile

Dataset/AUROC	Domain	Logistic Regression	Random Forest	XGBoost	Explainable Boosting Machine
Adult Income	Finance	.907±.003	.903±.002	.922±.002	.928±.002
Heart Disease	Medical	.895±.030	.890±.008	.870±.014	.916±.010
Breast Cancer	Medical	.995±.005	.992±.009	.995±.006	.995±.006
Telecom Churn	Business	.804±.015	.824±.002	.850±.006	.851±.005
Credit Fraud	Security	.979±.002	.950±.007	.981±.003	.975±.005

[Laurae2]

Note: EBM is a GLM with a quadratic design matrix, not a GBM. Not in the same category.

[szilard]

Thanks @mratsim for the info, I did not know about this lib. It's not GBM/GBDT as @Laurae2 was saying, but it's an interesting project for me for other reasons (it's based on Caruana's paper that I know and love https://www.microsoft.com/en-us/research/wp-content/uploads/2017/06/KDD2015FinalDraftIntelligibleModels4HealthCare_igt143e-caruanaA.pdf ). So I'll try out this (thanks again for posting here), but will not add the results to GBM-perf as this is not a GBM.

[paulbkoch]

I'm one of the InterpretML developers. It's true that the final output of the EBM algorithm is a GAM, however internally we use gradient boosted decision trees to build those GAMs. This is probably why they are competitive vs other gradient boosting machine packages.

More info at: http://www.cs.cornell.edu/~yinlou/papers/lou-kdd12.pdf

[szilard]

Thanks @paulbkoch