| Type: | Package |
| Title: | Purge Training Data from Models |
| Version: | 0.2.1 |
| Date: | 2015-10-29 |
| Maintainer: | Marc Maier <mmaier@massmutual.com> |
| Description: | Enables the removal of training data from fitted R models while retaining predict functionality. The purged models are more portable as their memory footprints do not scale with the training sample size. |
| License: | MIT + file LICENSE |
| LazyData: | TRUE |
| Imports: | methods |
| Suggests: | testthat, survival, lme4 (≥ 1.1-8), ranger, randomForest, rpart |
| NeedsCompilation: | no |
| Packaged: | 2017-02-12 02:08:55 UTC; mm90357 |
| Author: | Marc Maier [cre], Chaoqun Jia [ctb], MassMutual Advanced Analytics [aut] (http://datascience.massmutual.com) |
| RoxygenNote: | 6.0.1 |
| Repository: | CRAN |
| Date/Publication: | 2017-02-12 08:31:10 |
Purge training data from a model
Description
Most R model implementations store the training data within the fitted object, often many times. It can be useful to remove the embedded data for portability, especially if the only required functionality is to predict on new data.
Usage
purge(model)
## Default S3 method:
purge(model)
## S3 method for class 'glm'
purge(model)
## S3 method for class 'lm'
purge(model)
## S3 method for class 'merMod'
purge(model)
## S3 method for class 'glmerMod'
purge(model)
## S3 method for class 'rpart'
purge(model)
## S3 method for class 'randomForest'
purge(model)
## S3 method for class 'ranger'
purge(model)
## S3 method for class 'coxph'
purge(model)
Arguments
model |
A fitted R model object |
Value
A fitted R model object, purged of its training data, but retaining its predict functionality on new data
Methods (by class)
-
default: Default purge returns a copy of the model -
glm: Purges a glm model -
lm: Purges an lm model -
merMod: Purges a merMod, linear mixed-effects model -
glmerMod: Purges a glmerMod, generalized linear mixed-effects model -
rpart: Purges an rpart model -
randomForest: Purges a random forest model -
ranger: Purges a ranger model for classification, regression, or survival -
coxph: Purges a coxph model
Examples
x <- rnorm(1000)
y <- x + rnorm(1000)
unpurged.model <- lm(y ~ x)
purged.model <- purge(unpurged.model)
object.size(unpurged.model)
object.size(purged.model)