Some changes

# Clear workspace and graphs

if(!is.null(dev.list())) dev.off()

rm(list = ls()) 

library("caret")

library("data.table")

# Create the training and test datasets

set.seed(100)

load(file="01_data/biofilm.Rdata")

setDT(biofilm_data)

#Descriptive: There are several files in the dataset

file_list <- unique(biofilm_data$file)

# Let's get one out of it

onefile <- biofilm_data[file_list[2]==biofilm_data$file,]

# Let's graph the mz values against the counts

plot(onefile$mz,onefile$int,pch=16,type="h")

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

# Splitting based on outcome variables ----

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

## Step 1: Get row numbers for the training data ----

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

biofilm_data$mz <- round(biofilm_data$mz,digits=1)

setkeyv(biofilm_data,cols=c("file","indicator","mz"))

# This step takes a long time

biofilm_data[,.(int=sum(int)),

             by=c("file","indicator","mz")]

biofilm_data_wide <- dcast(biofilm_data, file + indicator ~ mz,value.var = "int")

biofilm_data_wide <- biofilm_data_wide[,-c("file")]

pcas <- prcomp(biofilm_data_wide)

trainRowNumbers <- createDataPartition(as.factor(biofilm_data_wide$indicator), p=0.8, list=FALSE,times=1)

# list: no list is returned

# times: you can do multiple times

# if the output is a factor splitting is done so that the the outcome variable is equally represented

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

## Step 2: Create the training  dataset ----

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

trainData <- biofilm_data_wide[trainRowNumbers,]

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

## Step 3: Create the test dataset ----

# +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

testData <- biofilm_data_wide[-trainRowNumbers,]

# What are the frequencies?

table(biofilm_data$indicator)/length(biofilm_data$indicator)

table(trainData$indicator)/length(trainData$indicator)

# The outcome variable Y for later use.

y = trainData$indicator