# BUAN 6356 Group Project
# Group Members: Akash Navneeth, Akshat Gadiya, & Rohith Thadela
# Project: Analyzing Attrition at IBM

# Importing the attrition_data.csv file
getwd()
setwd("C:/Users/akash/OneDrive/Desktop/UTD/Spring 2023 Classes/(5) BUAN 6356/Group Project")

attritiondata <- read.csv("attrition_data.csv", header=TRUE, stringsAsFactors = FALSE)

# Converting certain variables to factor type
attritiondata$Education <- as.factor(attritiondata$Education)
attritiondata$EnvironmentSatisfaction <- as.factor(attritiondata$EnvironmentSatisfaction)
attritiondata$JobSatisfaction <- as.factor(attritiondata$JobSatisfaction)
attritiondata$WorkLifeBalance <- as.factor(attritiondata$WorkLifeBalance)

# Loading the necessary libraries for the analysis
library(ggplot2)

# Insight 1: Do employees that travel a lot compared to other employees more likely to leave the company?
# Creating a histogram to display the data to answer this insight.

g1 <- ggplot(attritiondata, aes(x=DistanceFromHome, fill=Attrition)) + geom_histogram(color='black')
g1 <- g1 + labs(x='Distance From Home', y='Number of Employees', title='Does Distance From Home Impact Attrition?')
g1 <- g1 + facet_grid(vars(Attrition))
g1

# Insight 2: Are less educated people in the company more likely to leave the company?
# Creating a bar plot to display the data to answer this insight.

g2 <- ggplot(attritiondata, aes(x=Education, fill=Attrition)) + geom_bar(color='black', stat="count")
g2 <- g2 + labs(x='Level of Education', 'Number of Employees', title='Does Level of Education Impact Attrition?')
g2 <- g2 + facet_grid(vars(Attrition))
g2 <- g2 + scale_x_discrete(labels=c('Below College', 'College', 'Bachelors', 'Masters', 'Doctorate'))
g2

# Insight 3: Is there a specific department that seems to have a higher amount of people who decide to leave the company?
# Creating multiple pie charts to display the data to answer this insight.

g3 <- ggplot(attritiondata, aes(x=factor(1), stat='bin', fill=Attrition)) + geom_bar(position='fill')
g3 <- g3 + facet_grid(facets=. ~ Department) + coord_polar(theta="y")
g3 <- g3 + labs(x='', y='Department', title='Rate of Attrition By Department')
g3

# Insight 4: Do employees that have worked at a lot of companies prior to IBM have a higher chance of leaving the company?
# Creating a side-by-side bar chart to display the answer to this insight.

g4 <- ggplot(attritiondata, aes(x=NumCompaniesWorked, fill=Attrition))
g4 <- g4 + geom_bar(color='black', stat="count", position=position_dodge())
g4 <- g4 + scale_x_continuous(breaks=c(0,1,2,3,4,5,6,7,8,9))
g4 <- g4 + labs(x='Number of Companies Previously Worked', y='Number of Employees', title='Does Number of Companies Previously Worked Influence Attrition?')
g4

# Insight 5: Do employees that have worked for a long time with IBM have a higher chance of staying with the company?
# Creating a side-by-side bar chart to display the answer to this insight.

g5 <- ggplot(attritiondata, aes(x=YearsAtCompany, fill=Attrition))
g5 <- g5 + geom_bar(color='black', stat="count", position=position_dodge())
g5 <- g5 + labs(x='Tenure at IBM', y='Number of Employees', title='Does Tenure Impact Attrition?')
g5

# Insight 6: Which of the following factors has more of an impact on whether an employee might leave the company 
#            (and contribute to the attrition rate): job satisfaction, or environment satisfaction
#            or work-life balance?

# Creating a faceted pie chart to analyze Level of JOB SATISFACTION.
g6a <- ggplot(attritiondata, aes(x=factor(1), stat='bin', fill=Attrition)) + geom_bar(position='fill')
g6a <- g6a + facet_grid(facets=. ~ JobSatisfaction) + coord_polar(theta="y")
g6a <- g6a + labs(x='', y='Level of Job Satisfaction', title='Rate of Attrition By Level of Job Satisfaction')
g6a

# Creating a faceted pie chart to analyze Level of WORK-LIFE BALANCE.
g6b <- ggplot(attritiondata, aes(x=factor(1), stat='bin', fill=Attrition)) + geom_bar(position='fill')
g6b <- g6b + facet_grid(facets=. ~ WorkLifeBalance) + coord_polar(theta="y")
g6b <- g6b + labs(x='', y='Level of Work-Life Balance', title='Rate of Attrition By Level of Work-Life Balance')
g6b

# Creating a faceted pie chart to analyze Level of ENVIRONMENT SATISFACTION.
g6c <- ggplot(attritiondata, aes(x=factor(1), stat='bin', fill=Attrition)) + geom_bar(position='fill')
g6c <- g6c + facet_grid(facets=. ~ EnvironmentSatisfaction) + coord_polar(theta="y")
g6c <- g6c + labs(x='', y='Level of Environment Satisfaction', title='Rate of Attrition By Level of Environment Satisfaction')
g6c

# Insight 7: Is there a particular trend between age and monthly income received at IBM in relation to attrition?
# Creating a scatterplot with a trendline.

g7 <- ggplot(attritiondata, aes(x=Age, y=MonthlyIncome, color=Attrition))
g7 <- g7 + geom_point(stat="identity") + geom_smooth(method="loess")
g7 <- g7 + labs(y='Monthly Income', title='Is there a Correlation Between Age & Montly Income with Attrition?')
g7

# Building a Classification Model

set.seed(123)
index_train <- sample(1:nrow(attritiondata), 2 / 3 * nrow(attritiondata))
attritiondata.train <- attritiondata[index_train, ]
attritiondata.test <- attritiondata[-index_train, ]

# install.packages("rpart")
# install.packages("rpart.plot")

library(rpart)
library(rpart.plot)

fit <- rpart(Attrition ~ ., data = attritiondata.train, method = "class", control = rpart.control(xval = 10, minsplit = 50), parms = list(split = "gini"))

rpart.plot(fit, type = 1, extra = 1, main = "Classification Tree")

attritiondata.pred <- predict(fit, attritiondata.test, type="class")
attritiondata.actual <- attritiondata.test$Attrition

cm <- table(attritiondata.pred, attritiondata.actual)

tp <- cm[2,2]
tn <- cm[1,1]
fp <- cm[2,1]
fn <- cm[1,2]

# Calculation for False Positive Rate (FPR)
attritiondata.test.fpr <- (fp / (fp + tn))

# Calculation for False Negative Rate (FNR)
attritiondata.test.fnr <- (fn / (fn + tp))

# Calculation for Specificity
attritiondata.test.spec <- (tn / (tn + fp))

# Calculation for Sensitivity
attritiondata.test.sens <- (tp / (tp + fn))

# Calculation for Accuracy
attritiondata.test.acc <- (tp + tn) / (tp + tn + fp + fn)