library(ggplot2)
library(caret)
library(e1071)

data <- read.csv("RidingMowers.csv")

data1 <- subset(data, data$Ownership == "Owner")
data2 <- subset(data, data$Ownership == "Nonowner")

#A
(dim(data1)[1] / dim(data)[1]) * 100
summary(data$Ownership)

#B
ggplot(data = data, aes(x = data$Income, y = data$Lot_Size)) +
  geom_point(aes(col = factor(data$Ownership))) +
  theme_bw()

logit.reg <- glm(Ownership ~ ., data = data[, 1:3], family = "binomial")
summary(logit.reg)

pred <- predict(logit.reg, data2, type = "response")
valid.predict<-data.frame(actual = data2$Ownership, predicted = pred)
pred_rate <- subset(valid.predict, valid.predict$predicted < 0.5)

#C
(dim(pred_rate)[1] / dim(data2)[1]) * 100
confusionMatrix(as.factor(ifelse(valid.predict$predicted < 0.5, "Nonowner", "Owner")), data2$Ownership) #sensitivity


#E, F
new <- data.frame(Income = 60.0, Lot_Size = 20.0)
pred <- predict(logit.reg, new, type = "response") #nonowner
odds <- pred / (1 - pred)

#G
income_table <- data.frame(Income = seq(50,200,1), Probability = rep(0,151))

for (i in seq(50,200,1)){
  new_row <- data.frame(Income = i, Lot_Size = 16)
  pred <- predict(logit.reg, new_row, type = "response")
  income_table[income_table$Income == i, 2 ] <- pred
}