Slow mutate_if()

[mayer79]

The calculations of {randomForest} models lose a significant amount of time in mutate_if() calls. Replacing them by an explicit mutate() is easy, and gives a clear speed-up:

library(randomForest)
library(randomForestExplainer)

set.seed(12)

fit <- randomForest(Sepal.Width~., data = iris)
fit2 <- ranger(Sepal.Width~., data = iris)

system.time( # 2.1 seconds with mutate_if() -> 0.8 seconds with mutate()
out <- min_depth_distribution(fit)
)

[mayer79]

Done in #37