Abstract
We consider the problem of predicting an outcome variable on the basis of a small number of covariates, using an interpretable yet non-additive model. We propose convex regression with interpretable sharp partitions (CRISP) for this task. CRISP partitions the covariate space into blocks in a data-adaptive way, and fits a mean model within each block. Unlike other partitioning methods, CRISP is fit using a non-greedy approach by solving a convex optimization problem, resulting in low-variance fits. We explore the properties of CRISP, and evaluate its performance in a simulation study and on a housing price data set.
Original language | English (US) |
---|---|
Pages (from-to) | 1-31 |
Number of pages | 31 |
Journal | Journal of Machine Learning Research |
Volume | 17 |
State | Published - Jun 1 2016 |
Bibliographical note
Funding Information:We thank the associate editor and three referees for helpful comments. D.W. was supported by NIH Grant DP5OD009145, NSF CAREER Award DMS-1252624, and an Alfred P. Sloan Foundation Research Fellowship. N.S. was supported by NIH Grant DP5OD019820.
Publisher Copyright:
©2016 Ashley Petersen, Noah Simon, and Daniela Witten.
Keywords
- Convex optimization
- Interpretability
- Non-additivity
- Non-parametric regression
- Prediction