tweak: Calcurate suggention for each inputs based on prediction

In katokohaku/featureTweakR: Calculate Actionable Feature Tweaking

Description

Calcurate suggention for each inputs based on prediction

Usage

tweak(esrules, forest, newdata, label.from, label.to, .dopar = TRUE)

Arguments

esrules	an object returned by set.eSatisfactory().
forest	an object of ensemble trees to be parsed. See randomForest or xgboost.
newdata	a data.frame or matrix. test data to be predicted and to be suggested. newdata must have the same structure of train data.
label.from	a character. Predicted class label that user wants to change.
label.to	a character. Class label that user wants to be changed from label.from.
.dopar	logical. If .dopar = TRUE, suggestion for each instance will be calculated in parallel

Value

a list of

• "predict" character vector of predicted label of each instance. See predict.randomForest.

• "original" is the same as newdata.

• "suggest" a data.frame of instances tweakd from newdata.

Examples

## Not run: 
X <- iris[, 1:(ncol(iris)-1)]
scaled.X <- scale(X)
true.y <- iris[, ncol(iris)]

rf.iris <- randomForest(X, true.y, ntree=30)
rules.rf <- getRules(rf.iris, ktree = 20)
es.rf <- set.eSatisfactory(forest.rf, ntree = 30, epsiron = 0.3, resample = TRUE)
tweaked <- tweak(es.rf, newdata= scaled.X, label.from = "spam", label.to = "nonspam",
                 .dopar = TRUE)

## End(Not run)

katokohaku/featureTweakR documentation built on May 17, 2019, 11:17 p.m.