bbl (version 0.1.5)

accuracy: Compute Prediction Accuracy

Description

Accuracy of predicted response probability is computed.

Usage

accuracy(object, prediction, balanced = FALSE)

Arguments

object

Object of class bbl with test data in data slot.

prediction

Data frame of predicted response group probability from predict.

balanced

Compute balanced accuracy. If TRUE, $$s = \frac{1}{K}\sum_y \frac{1}{n_y} \sum_{k\in y} \delta\left({\hat y}_k = y\right).$$ If FALSE, $$s = \frac{1}{n}\sum_{k} \delta\left({\hat y}_k = y_k\right).$$

Value

List of acc (accuracy score) and yhat (predicted response group).

Details

An option is provided for computing group-balanced accuracy, where prediction score is calculated for each group separately and averaged.

Examples

# NOT RUN {
titanic <- freq2raw(as.data.frame(Titanic), Freq='Freq')
nsample <- NROW(titanic)
mod <- bbl(data=titanic, y='Survived')
mod <- mod[sample(nsample),]
mtrain <- mod[seq(nsample/2),]
mtest <- mod[seq(nsample/2,nsample),]
mtrain <- train(mtrain, method='mf')
pred <- predict(mtrain, newdata=mtest@data)
score <- accuracy(mtest, prediction=pred, balanced=TRUE) 
# }