Random Forest

Random Forest

# soils
# set this to the xlsx file on your computer
setwd("c:/Users/machariam/Desktop/Soil_R_project") # Change working directory
# load libraries
lapply(c("data.table", "DescTools", "stringr", "ggplot2", "readxl", "ranger", "RWeka", "Boruta", "DMwR"), require, character.only = T)
# load data
myfiles = dir(pattern = "xlsx") #These functions produce a character vector of the names of files or directories in the named directory.
myfiles #See no value in this unless it can be linked with the read_excel below
df = read_excel(myfiles, na = "") # Load the excel, still need a way to sort out the data structure and missing values
df
setDT(df)#When working on large lists or data.frames, it might be both time and memory consuming to convert them to a data.table using as.data.table(.), as this will make a complete copy of the input object before to convert it to a data.table. The setDT function takes care of this issue by allowing to convert lists - both named and unnamed lists and data.frames by reference instead. That is, the input object is modified in place, no copy is being made.
str(df)
names(df)
# select some variables
m = df[ , .(Nrate,Prate,Krate,Pppm,ExchK,ExchCa,ExchMg,ExchNa,ExchAl,CEC,ECEC,Sand,Silt,Clay,GrainYield)]
m
str(m)
# plot missing values and impute
PlotMiss(m, clust = F)# {DescTools}Takes a data.frame or data.table and displays the location of missing data. The missings can be clustered and be displayed together. ??PlotMiss
m = knnImputation(m) #Not sufficient complete cases for computing neighbors.
View(m)
library("VIM")
kNN(m, k=3)

#More on imputation
#Try MICE package

require(mice)
pMiss <- function(m){sum(is.na(m))/length(m)*100}
pMiss 
apply(m,2,pMiss)
apply(m,1,pMiss)
library(VIM)
aggr_plot <- aggr(m, col=c('navyblue','red'), numbers=TRUE, sortVars=TRUE, labels=names(m), cex.axis=.7, gap=3, ylab=c("Histogram of missing data","Pattern"))


# which variables are important?
Boruta(GrainYield ~., m, doTrace = 2)
# model
r = ranger(grainyield ~., m, importance = "impurity", write.forest = T)
r$variable.importance %>% PlotDot
pred = predict(r, m)

# plot
p = data.table(yield = m$grainyield, pred.yield = pred$predictions)
cor(m$grainyield, pred$predictions)

m[, var := df$variety1]
m[is.na(var), var := "unknown"]

# what are the frequency distributions of the variables?
# (three guesses: *not* normal ;-))
Desc(m, plot = T)