# Telco Customer Churn Analysis Script

# Load required libraries

library(ggplot2)
library(dplyr)
library(rpart)
library(e1071)
library(caret)
library(pROC)

# Load dataset

telco <- read.csv("Telco-Customer-Churn.csv", stringsAsFactors = TRUE)
telco\$TotalCharges <- as.numeric(as.character(telco\$TotalCharges))
telco <- telco\[!is.na(telco\$TotalCharges), ]
telco\$Churn <- factor(telco\$Churn, levels = c("No", "Yes"))

# Exploratory Data Visualizations

# Histogram for numeric variables

numeric\_vars <- c("tenure", "MonthlyCharges", "TotalCharges")
for (var in numeric\_vars) {
p <- ggplot(telco, aes\_string(x = var)) +
geom\_histogram(binwidth = 10, fill = "skyblue", color = "black") +
labs(title = paste("Histogram of", var), x = var, y = "Frequency") +
theme\_minimal()
ggsave(paste0("hist\_", var, ".png"), plot = p)
}

# Bar plot for churn

p\_churn <- ggplot(telco, aes(x = Churn, fill = Churn)) +
geom\_bar() +
labs(title = "Churn Distribution", x = "Churn", y = "Count") +
theme\_minimal()
ggsave("bar\_churn.png", plot = p\_churn)

# Boxplot of MonthlyCharges by Churn

p\_box <- ggplot(telco, aes(x = Churn, y = MonthlyCharges, fill = Churn)) +
geom\_boxplot() +
labs(title = "Monthly Charges by Churn", x = "Churn", y = "Monthly Charges") +
theme\_minimal()
ggsave("boxplot\_monthlycharges\_churn.png", plot = p\_box)

# Split data into training, validation1, and validation2

set.seed(100)
n <- nrow(telco)
train.index <- sample(1\:n, size = round(0.70 \* n))
remaining.index <- setdiff(1\:n, train.index)
valid1.index <- sample(remaining.index, size = round(0.15 \* n))
valid2.index <- setdiff(remaining.index, valid1.index)

train.df <- telco\[train.index, ]
valid1.df <- telco\[valid1.index, ]
valid2.df <- telco\[valid2.index, ]

# Logistic regression model (simplified)

logit.reg <- glm(Churn \~ SeniorCitizen + Dependents + tenure + MultipleLines + InternetService + Contract +
PaperlessBilling + PaymentMethod + MonthlyCharges + TotalCharges,
data = train.df, family = "binomial")
summary(logit.reg)

# Evaluate on validation set

valid1\_pred\_probs <- predict(logit.reg, newdata = valid1.df, type = "response")
valid1\_pred <- factor(ifelse(valid1\_pred\_probs > 0.5, "Yes", "No"), levels = c("No", "Yes"))
logit\_conf <- confusionMatrix(valid1\_pred, valid1.df\$Churn, positive = "Yes")
logit\_roc <- roc(response = valid1.df\$Churn, predictor = valid1\_pred\_probs)

# Decision Tree model

dt\_model <- rpart(Churn \~ tenure + MonthlyCharges + TotalCharges + SeniorCitizen, data = train.df, method = "class")
dt\_pred <- predict(dt\_model, valid1.df, type = "class")
dt\_conf <- confusionMatrix(dt\_pred, valid1.df\$Churn)

# Naive Bayes model

nb\_model <- naiveBayes(Churn \~ tenure + MonthlyCharges + TotalCharges + SeniorCitizen, data = train.df)
nb\_pred <- predict(nb\_model, valid1.df)
nb\_probs <- predict(nb\_model, valid1.df, type = "raw")
nb\_conf <- confusionMatrix(nb\_pred, valid1.df\$Churn)
nb\_roc <- roc(response = valid1.df\$Churn, predictor = nb\_probs\[,"Yes"])

# Print evaluations

cat("\nLogistic Regression Confusion Matrix:\n")
print(logit\_conf)
cat("\nAUC (Logistic):", auc(logit\_roc), "\n")

cat("\nDecision Tree Confusion Matrix:\n")
print(dt\_conf)

cat("\nNaive Bayes Confusion Matrix:\n")
print(nb\_conf)
cat("\nAUC (Naive Bayes):", auc(nb\_roc), "\n")

# Save ROC curve plots

png("logistic\_roc\_curve.png")
plot(logit\_roc, main = "ROC Curve - Logistic Regression", col = "darkgreen")
dev.off()

png("naive\_bayes\_roc\_curve.png")
plot(nb\_roc, main = "ROC Curve - Naive Bayes", col = "blue")
dev.off()