blog example adaStump.R

library(adaStump)

#Load Letters Dataset

letters.data <- read.csv("https://archive.ics.uci.edu/ml/machine-learning-databases/letter-recognition/letter-recognition.data", header = F)

names(letters.data) <- c("lettr", "xbox", "ybox", "width", 
                          "high", "onpix", "xbar", "ybar",
                          "x2bar", "y2bar", "xybar", "x2ybr",
                         "xy2br", "xege", "xegvy", "yege", "yegvx")

#Create a binary clas

letters.data$isA <- ifelse(letters.data$lettr == "A","Yes","No")

#Create model formula. Exclude letter var for obvious reasons

antec <- paste0(setdiff(names(letters.data),c("isA","lettr")), collapse = "+")

fla <- as.formula(paste("isA", antec, sep = "~"))

#### Model training comparison ####

#ada model fit passing the corresponding control function to create Stumps

fit_with.ada <- ada(formula = fla,data = letters.data, type = "real",
    control = rpart.control(maxdepth=1,cp=-1,minsplit=0,xval=0), iter = 40, nu = 0.05, bag.frac = 1)

#adaStump

fit_with.adaStump <- adaStump(formula = fla,data = letters.data, 
                              type = "real", iter = 40, nu = 0.05, bag.frac = 1)

#Object size comparison

format(object.size(fit_with.ada), units = "KB")
format(object.size(fit_with.adaStump), units = "KB")

#### Model testing ####

system.time(predictions_ada <- predict(fit_with.ada, letters.data, type = "prob"))

system.time(predictions_adaStump <- predict(fit_with.adaStump, letters.data))

#### Model pruning ####

fit_with.prunedadaStump <- pruneTree(fit_with.adaStump)
system.time(predictions_prunedadaStump <- predict(fit_with.prunedadaStump, letters.data))