##########
#Example 4 (using the tidymodels library)
##########
##########
### in case we haven't installed the necessary libraries
# install.packages("tidymodels")
# install.packages("ggplot2")

# Load necessary libraries
library(tidymodels)
library(rpart.plot)
library(ggplot2)

# Set seed for reproducibility
set.seed(123)

# Load the dataset
data(iris)

# Split into training and testing sets (80/20 split)
iris_split <- initial_split(iris, prop = 0.8)
iris_train <- training(iris_split)
iris_test  <- testing(iris_split)

# Define the decision tree model specification
dt_spec <- decision_tree() %>%
  set_engine("rpart") %>%
  set_mode("classification")

# Create the workflow
iris_wf <- workflow() %>%
  add_model(dt_spec) %>%
  add_formula(Species ~ .)

# Fit the model
iris_fit <- fit(iris_wf, data = iris_train)

# Extract the fitted rpart object for visualization
rpart_obj <- pull_workflow_fit(iris_fit)$fit

# Plot the decision tree
rpart.plot(rpart_obj, main = "Decision Tree for Iris (tidymodels)")

# Make predictions on the test set
iris_preds <- predict(iris_fit, iris_test) %>%
  bind_cols(iris_test)

# Show first few predictions
print(head(iris_preds))

# Generate confusion matrix
conf_mat_obj <- conf_mat(iris_preds, truth = Species, estimate = .pred_class)

# Plot confusion matrix
autoplot(conf_mat_obj) + ggtitle("Confusion Matrix for Iris Classification")

# Calculate accuracy
accuracy_val <- accuracy(iris_preds, truth = Species, estimate = .pred_class)
print(accuracy_val)

# Visualize misclassified points
ggplot(iris_preds, aes(x = Sepal.Length, y = Sepal.Width, 
                       color = Species, shape = .pred_class)) +
  geom_point(size=3) +
  labs(title = "Misclassified Points in Iris Dataset",
       shape = "Predicted Class") +
  theme_minimal()