GBM Heart Disease

library(dplyr)
library(class)
library(gmodels)
library(caret)
library(gridExtra)

setwd("C:/Users/tyler/Desktop/DATA FILES")
#read in data
heart <- read.csv("heart.csv", stringsAsFactors=FALSE)

#reclassify feature as a factor and add labels (required for ML)
heart$target <- factor(heart$target, levels = c(0,1),
                       labels = c("No", "Yes"))

#percent yes / no
round(prop.table(table(heart$target))*100, digits = 1)

#summarize 3 features
summary(heart[c(colnames(heart))])
#features have different scales need to normalize

#normalize function to normalize all data between 0 and 1
norm <- function(x){
  return((x-min(x))/(max(x)-min(x)))
}

#test the norm function
norm(c(1,2,3,4,5))
norm(c(100,200,300,400,500))

#normalize all of the columns in the df using lappy (skip col 1)
heart_norm <- as.data.frame(lapply(heart[1:13], norm))

heart_norm_full <- cbind(heart_norm, heart[14])

#check summary of normalized data
summary(heart_norm[c(colnames(heart_norm))])

#split train test
set.seed(123)
train.index <- createDataPartition(heart$target, times = 1, p = 0.80, list = FALSE)
heart_train <- heart_norm_full[train.index,]
heart_test <-  heart_norm_full[-train.index,]


#plot accuracy of GBMS based on 3, 5, 10, and 20 n.minobsinnode
# Create a list to hold the plot objects.
pltList <- list()
vec <- c(3,5,10,20)
for (i in vec) {
  gbmGrid <-  expand.grid(interaction.depth = c(1, 5, 9), 
                          n.trees = (1:30)*50, 
                          shrinkage = 0.1,
                          n.minobsinnode = i)
  set.seed(321)
  gbmfit <- train(target ~ ., data = heart_train, 
                  method = "gbm", 
                  trControl = trainControl(method = "cv"),
                  verbose = FALSE,
                  tuneGrid = gbmGrid)
  #set theme and plot
  # Create plot name.
  pltName <- paste( 'gbm_stop', i, sep = '' )
  trellis.par.set(caretTheme())
  pltList[[ pltName ]] <- ggplot(gbmfit)+ labs(title = paste("minobsinnode", i))
}

#arrange plots
n <- length(pltList)
nCol <- floor(sqrt(n))
do.call("grid.arrange", c(pltList, ncol=nCol))