DEWScontrol: Set up parameters to control the modelSearch and modelBuild...

In jknowles/ModelEWS: Wisconsin Dropout Early Warning System

Description

This function defines three possible streams for the DEWS model searching and model training routines to take. PROD is the most rigorous, while DEV is the quickest for debugging and testing purposes. TEST is somewhere in the middle. Here the

Usage

DEWScontrol(mode, grade = NULL)

Arguments

mode	character vector with three choices - "DEV", "PROD", and "TEST"
grade	a numeric specifying the grade level

Value

A list with elements

• upsample - a character representing the upsampling technique to use, currently only SMOTE is supported

• length - a numeric representing the length of the tuneGrid to be passed to the tuneLength option in caret

• metric - a character representing the metric to use to identify the best model, currently ROC is all that is supported

• core - a numeric specifying the number of cores to use. In DEV, this is capped at 1/2 the maximum number of available cores, in other modes it is 2/3

• folds - a numeric representing the number of folds to use in training models

• reps - numeric representing the number of repeats to use in training models

• cvmethod - character representing the method to use for cross-validation, PROD is repeatedcv, dev and test are cv

• dewsTrControl - a trainControl object derived from values

• mode - a character storing the mode

Any of these values can be modified posthoc by editing the list elements directly.

Note

This function is extremely proscriptive and should be modified based on the user environment.

jknowles/ModelEWS documentation built on May 19, 2019, 11:42 a.m.