Housing_and_Phishing_Data_Analysis.Rmd

---
title: "Housing and Phishing Data Analysis"
output: html_document
---

## Introduction

This document contains an analysis of housing and phishing datasets.

## Load Libraries

```{r}
library(randomForest)
library(ggplot2)
library(e1071)
library(caTools)
library(caret)
library(rpart)
library(foreign)

## Housing 

housing_data <- read.csv("/file/path")

na_count <- sapply(housing_data, function(y) sum(is.na(y)))
print(paste("NA counts:", na_count))
housing_data <- na.omit(housing_data)

ggplot(housing_data, aes(x = longitude, y = latitude)) +
  geom_point(aes(color = median_house_value, size = population), alpha = 0.4) +
  labs(title = "California Housing Prices", x = "Longitude", y = "Latitude")

set.seed(123)
samp <- sample(1:nrow(housing_data), 0.7 * nrow(housing_data))
train_data <- housing_data[samp, ]
test_data <- housing_data[-samp, ]

rf_model <- randomForest(median_house_value ~ ., data = train_data, ntree = 100)
pred <- predict(rf_model, newdata = test_data)
mse <- mean((pred - test_data$median_house_value)^2)
print(paste("Test MSE: ", round(mse, 2)))


## problem 4: positive class = when the loan is paid back by an individual
## recall 75 = model predicted accurately 75% of all actual positive cases 
# recall 85 = of all predicted positive cases, 85% actually did pay back the loan
## choose accuracy 0.967


## Phishing 

phishing_data <- read.arff("/file/path")
print(sum(is.na(phishing_data)))
phishing_data <- na.omit(phishing_data)

set.seed(123)
split <- sample.split(phishing_data$Result, SplitRatio = 0.8)
train_data_phishing <- subset(phishing_data, split == TRUE)
test_data_phishing <- subset(phishing_data, split == FALSE)

rf_model_phishing <- randomForest(Result ~ ., data = train_data_phishing)
predicted_values_rf <- predict(rf_model_phishing, newdata = test_data_phishing)
cm_rf <- confusionMatrix(predicted_values_rf, test_data_phishing$Result)
print(cm_rf)

dt_model <- rpart(Result ~ ., data = train_data_phishing, method = "class")

predicted_values_dt <- predict(dt_model, newdata = test_data_phishing, type = "class")
cm_dt <- confusionMatrix(predicted_values_dt, test_data_phishing$Result)
print(cm_dt)

## random forest has higher accuracy , specificity, sensitivity, ppv, npv, and kappa
## random forest also has higher accuracy


```