predict.modelBst: Predict Method for modelBst Object

In ebmc: Ensemble-Based Methods for Class Imbalance Problem

Description

Predicting instances in test set using modelBst object

Usage

  ## S3 method for class 'modelBst'
predict(object, newdata, type = "prob", ...)

Arguments

object	A object of modelBst class.
newdata	A data frame object containing new instances.
type	Types of output, which can be prob (probability) and class (predicted label). Default is prob.
...	Not used currently.

Value

Two type of output can be selected:

prob	Estimated probability of being a minority instance (i.e. 1). The probability is averaged by using a majority vote by all weak learners, weighted by error estimation.
class	Predicted class of the instance. Instances of probability larger than 0.5 are predicted as 1, otherwise 0.

Examples

data("iris")
iris <- iris[1:70, ]
iris$Species <- factor(iris$Species, levels = c("setosa", "versicolor"), labels = c("0", "1"))
samp <- sample(nrow(iris), nrow(iris) * 0.7)
train <- iris[samp, ]
test <- iris[-samp, ]
model <- rus(Species ~ ., data = train, size = 10, alg = "c50") # Build RUSBoost model
prob <- predict(model, newdata = test, type = "prob") # return probability estimation
pred <- predict(model, newdata = test, type = "class") # return predicted class

Example output

ebmc documentation built on Jan. 11, 2022, 1:06 a.m.