# R Script for Logistic Regression

# --------------------------------------------------------------------------
# I. SETUP: Install and load necessary packages
# --------------------------------------------------------------------------

# This function checks if a package is installed. If not, it installs it.
# Then, it loads the package into the R session.
install_and_load <- function(package) {
  if (!require(package, character.only = TRUE)) {
    install.packages(package, dependencies = TRUE)
    library(package, character.only = TRUE)
  }
}

# List of packages to install and load
packages <- c("tidyverse", "caret", "caTools")
lapply(packages, install_and_load)

# --------------------------------------------------------------------------
# II. DATA IMPORT: Read the CSV file
# --------------------------------------------------------------------------

# IMPORTANT: Replace "path/to/your/file.csv" with the actual path to your CSV file.
# For example, "C:/Users/YourName/Documents/data.csv" on Windows or
# "/home/YourName/data/data.csv" on Linux/macOS.
# If the file is in your current working directory, you can just use "file.csv".

tryCatch({
  # Read the data from the CSV file
  # The `read_csv` function from the `readr` package (part of tidyverse) is used here.
  df <- read_csv("path/to/your/file.csv")
  
  # Display the first few rows of the dataframe to inspect it
  print("Data successfully loaded. Here are the first 6 rows:")
  print(head(df))
  
  # Display the structure of the dataframe to see column types
  print("Structure of the dataframe:")
  str(df)
  
}, error = function(e) {
  # This message will be displayed if the file is not found or another error occurs.
  message("Error reading the CSV file. Please check the following:")
  message("- The file path is correct.")
  message("- The file is a valid CSV.")
  message("- You have the necessary permissions to read the file.")
  message("Original error message:")
  message(e)
})


# --------------------------------------------------------------------------
# III. DATA PREPROCESSING
# --------------------------------------------------------------------------

# This section assumes your dataframe `df` is now loaded.
if (exists("df")) {

  # 1. Handle Missing Values (NA)
  # -----------------------------
  # Check for missing values in each column
  print("Count of missing values in each column:")
  print(sapply(df, function(x) sum(is.na(x))))

  # Option A: Remove rows with any missing values.
  # This is the simplest method but can lead to data loss.
  # df <- na.omit(df) 
  
  # Option B: Impute missing values.
  # For numeric columns, we can replace NA with the mean or median.
  # For categorical columns, we can replace NA with the mode.
  df <- df %>%
    mutate(across(where(is.numeric), ~ifelse(is.na(.), mean(., na.rm = TRUE), .))) %>%
    mutate(across(where(is.character), ~ifelse(is.na(.), names(which.max(table(.))), .))) %>%
    mutate(across(where(is.factor), ~ifelse(is.na(.), names(which.max(table(.))), .)))

  print("Missing values have been imputed.")
  print(sapply(df, function(x) sum(is.na(x))))


  # 2. Encode Categorical Variables
  # ---------------------------------
  # Logistic regression requires the dependent variable to be a factor.
  # Replace 'YourDependentVariable' with the name of your outcome column.
  # The dependent variable should be binary (have only two levels, e.g., "Yes"/"No", 0/1).
  
  # Example: If your dependent variable is named 'churn' with values "Yes" and "No"
  # df$churn <- as.factor(df$churn)
  
  # Make sure to replace 'YourDependentVariable' with your actual column name
  # dependent_variable_name <- "YourDependentVariable"
  # df[[dependent_variable_name]] <- as.factor(df[[dependent_variable_name]])
  
  # It's also good practice to convert character predictors to factors
  df <- df %>%
    mutate(across(where(is.character), as.factor))

  print("Data types after encoding:")
  str(df)

  # 3. Data Splitting: Training and Testing Sets
  # ----------------------------------------------
  # We'll split the data to train the model and then test its performance on unseen data.
  # Replace 'YourDependentVariable' with the name of your outcome column.
  
  # set.seed(123) # for reproducibility
  # sample_split <- sample.split(df$YourDependentVariable, SplitRatio = 0.80)
  # train_set <- subset(df, sample_split == TRUE)
  # test_set <- subset(df, sample_split == FALSE)
  
  # print(paste("Training set size:", nrow(train_set)))
  # print(paste("Test set size:", nrow(test_set)))

}

# --------------------------------------------------------------------------
# IV. LOGISTIC REGRESSION MODEL
# --------------------------------------------------------------------------

# This section builds the logistic regression model.
# Make sure you have defined `train_set` and `test_set` above.
# Replace 'YourDependentVariable' with your actual dependent variable name.

if (exists("train_set") && exists("test_set")) {

  # 1. Build the Model
  # --------------------
  # The formula `YourDependentVariable ~ .` means we are predicting the dependent variable
  # using all other columns as predictors.
  # You can also specify predictors manually, e.g., `YourDependentVariable ~ predictor1 + predictor2`
  
  # logistic_model <- glm(YourDependentVariable ~ ., 
  #                       data = train_set, 
  #                       family = binomial)
  
  # print("Logistic Regression Model Summary:")
  # print(summary(logistic_model))

  # 2. Make Predictions
  # -------------------
  # We use the trained model to predict probabilities on the test set.
  # `type = "response"` gives probabilities P(Y=1).
  
  # predicted_probs <- predict(logistic_model, newdata = test_set, type = "response")
  
  # Convert probabilities to class predictions ("0" or "1", "Yes" or "No")
  # We use a threshold of 0.5. If probability > 0.5, predict 1 (or "Yes"), otherwise 0 (or "No").
  # predicted_classes <- ifelse(predicted_probs > 0.5, 1, 0) # Use 1/0 or "Yes"/"No" depending on your factor levels
  
  # 3. Evaluate Model Performance
  # -----------------------------
  # Create a confusion matrix to see how well the model performed.
  # Make sure the levels of your predicted classes and actual values match.
  
  # actual_values <- test_set$YourDependentVariable
  # confusion_matrix <- table(Predicted = predicted_classes, Actual = actual_values)
  
  # print("Confusion Matrix:")
  # print(confusion_matrix)
  
  # Calculate accuracy
  # accuracy <- sum(diag(confusion_matrix)) / sum(confusion_matrix)
  # print(paste("Model Accuracy:", round(accuracy, 4)))
  
  # Using caret for more detailed metrics
  # confusionMatrix_caret <- confusionMatrix(data = as.factor(predicted_classes), 
  #                                          reference = as.factor(actual_values))
  # print(confusionMatrix_caret)

} else {
  message("\nSkipping Model Building: 'train_set' or 'test_set' not found.")
  message("Please uncomment and run the data splitting section (III.3) first.")
  message("Remember to replace 'YourDependentVariable' with your actual target column name.")
}