# ==============================================================================
# --------------------------------- HOMEWORK 1 ---------------------------------
# ==============================================================================


library(MASS)
library(corrplot)


# ----------------------------------- TASK 1 -----------------------------------


load("customer_behaviour.RData")

data <- data[data$money_spent > 0, ]

data <- data[data$age > 0, ]

data <- data[data$web_visits >= 0, ]
data$web_visits <- as.numeric(data$web_visits)
data <- data[!is.na(data$web_visits), ]

data <- data[data$mail_ads >= 0, ]
data <- data[data$mail_ads < 1000, ]

data$shop_visits <- as.numeric(data$shop_visits)
data <- data[!is.na(data$shop_visits), ]

data <- data[data$shop_visits >= 0, ]


# ----------------------------------- TASK 2 -----------------------------------


data$big = (data[, 1] > 5000) + 2

boxplot(data$money_spent, col = "yellow")
boxplot(data$money_spent ~ data$big, col = 2:3,
        xlab = "big money indicator", ylab = "money spent", main = "boxplot")

d <- density(data$money_spent)
hist(data$money_spent, col = "red", freq = F, xlab = "money spent", main = "histogram")
box()
lines(d, col = "black", lwd = 2)

mean(data$age)
quantile(data$age, probs = c(0.25, 0.5, 0.75))
diff(quantile(data$age, probs = c(0.25, 0.75))) # IQR
var(data$age)


# ----------------------------------- TASK 3 -----------------------------------


est_cor <- cor(data)
sum_diag <- sum(diag(est_cor))

corrplot(est_cor, method = "circle")

cor(data$money_spent, data$age) 
cor(data$money_spent, data$web_visits)
cor(data$money_spent, data$shop_visits)
cor(data$money_spent, data$mail_ads)


# ----------------------------------- TASK 4 -----------------------------------


data.pca <- prcomp(data[, 2:5], scale = T)

(s <- summary(data.pca))
s$rotation

library(ggfortify)

autoplot(data.pca, data = data[, 1:5], colour = data$big)
autoplot(data.pca, data = data[, 1:5], colour = data$big, loadings = T, loadings.label = T)