Preprint Open Access

vtreat: a data.frame Processor for Predictive Modeling

Zumel, Nina; Mount, John

We look at common problems found in data that is used for predictive modeling tasks, and describe how to address them with the vtreat R package. vtreat prepares real-world data for predictive modeling in a reproducible and statistically sound manner. We describe the theory of preparing variables so that data has fewer exceptional cases, making it easier to safely use models in production. Common problems dealt with include: infinite values, invalid values, NA, too many categorical levels, rare categorical levels, and new categorical levels (levels seen during application, but not during training). Of special interest are techniques needed to avoid needlessly introducing undesirable nested modeling bias (which is a risk when using a data-preprocessor).

Files (547.9 kB)
Name Size
vtreat.pdf
md5:12877eceb9f09ea567b3d9f884905d7c
547.9 kB Download
49
4
views
downloads
All versions This version
Views 4949
Downloads 44
Data volume 2.2 MB2.2 MB
Unique views 4949
Unique downloads 44

Share

Cite as