BenHeubl

model_random_forrest_optimal <- randomForest(INCOME ~ ., 
data = TrainSet,
ntree = 500, mtry = 3, 
importance = TRUE)

model_decision_RF = predict(model_random_forrest_optimal, data = TrainSet)
table(model_decision_RF, TrainSet$INCOME)
mean(model_decision_RF == TrainSet$INCOME)

#[1] 0.7757143

BenHeubl

install.packages("rpart")
install.packages("caret")
install.packages("e1071")
library(rpart)
library(caret)
library(e1071)

model_decision_tree = train(INCOME ~ ., data = TrainSet, method = "rpart")

model_decision_tree_prediction = predict(model_decision_tree, data = TrainSet)

BenHeubl

# Using For loop to identify the right mtry for model (this took around 4 minutes for me. Get yourself a drink :-)

accuracy_list =c()
for (i in 3:8) {
print(i)
model_optimal <- randomForest(INCOME ~ ., data = TrainSet, ntree = 500, mtry = i, importance = TRUE)
predValid <- predict(model_optimal, ValidSet, type = "class")
accuracy_list[i-2] = mean(predValid == ValidSet$INCOME)
}

BenHeubl

income <- read.csv("https://raw.githubusercontent.com/selva86/datasets/master/income.csv")

incomeR <- income %>%
mutate(INCOME = if_else(INCOME == "-10.000)", "Under 30k",
if_else(INCOME == "[10.000–15.000)", "Under 30k",
if_else(INCOME == "[15.000–20.000)", "Under 30k",
if_else(INCOME == "[20.000–25.000)", "Under 30k",
if_else(INCOME == "[25.000–30.000)", "Under 30k", 'Over 30k')))))) %>% mutate_if(is.factor, fct_explicit_na, na_level = 'Unknown') %>%
mutate(INCOME = as.factor(INCOME))

BenHeubl

#mode function
getmode <- function(v) {
 uniqv <- unique(v)
 uniqv[which.max(tabulate(match(v, uniqv)))]
}

incomeR_mode_income <- incomeR %>% 
 group_by(INCOME) %>%
 summarise(mode = getmode(OCCUPATION))
 

BenHeubl

# ggplotting our featuer importance:
Feature_importance <- importance(model_base)
var_Importance <- data.frame(Variables = row.names(Feature_importance), 
Importance = round(importance[ ,'MeanDecreaseGini'],2))

#Create ranks for variable based on importance
Rank_Importance <- var_Importance %>%
mutate(Rank = paste0('#',dense_rank(desc(Importance))))

#Relative importance of our varaibles

BenHeubl

# read in the data from github repo:
income <- read.csv("https://raw.githubusercontent.com/selva86/datasets/master/income.csv")

set.seed(100)

# We shuffle row-wise:
incomeR <- income[sample(nrow(income)),]

#check rownames (see above screenshot)
colnames(incomeR)

BenHeubl

install.packages("relaimpo")
library(relaimpo)

#fit linear model:
Ozone_model <- lm(ozone_reading ~ . , data = Ozone)

#Get relative importance:
Relative_importance <- calc.relimp(lmMod, type = "lmg", rela = TRUE)

# Relative importance scaled to 100 and plot:

BenHeubl

install.packages('PerformanceAnalytics')
library(PerformanceAnalytics)

chart.Correlation(Ozone, histogram=TRUE, pch=19)

BenHeubl

#read in data
Ozone <- read.csv("https://raw.githubusercontent.com/selva86/datasets/master/ozone.csv", stringsAsFactors=F)