DA5030 Naive Bayes

Question 2

# URL of the CSV file
url <- "https://s3.us-east-2.amazonaws.com/artificium.us/datasets/HouseVotes-1984-ButOne.csv"

# Read the CSV file into a data frame
df <- read.csv(url, stringsAsFactors = FALSE, header = T)

# Show the first five rows of df
head(df, 5)

Question 3

# Empty lists for training and validation data
training_data <- list()
validation_data <- list()

set.seed(100)

# 5 random samples
for (i in 1:5) {
  # Sample 350 rows
  train_indices <- sample(nrow(df), 350)
  
  # Training data
  training_data[[i]] <- df[train_indices, ]
  
  # Validation data
  validation_data[[i]] <- df[-train_indices, ]
}

Question 4

# Empty list to store the models
nb_models <- list()

# Train Naive Bayes models 
nb_models <- lapply(training_data, function(train_set) {
  # Convert the party column to a factor
  train_set$party <- as.factor(train_set$Class)
  
  # Train the Naive Bayes model
  model <- naiveBayes(Class ~ ., data = train_set)
  
  return(model)
})

Question 5

# Ensemble function
predict_party <- function(models, new_cases) {
  # Vector to store predictions
  predictions <- vector("list", length(models))
  
  # Make predictions
  for (i in seq_along(models)) {
    predictions[[i]] <- predict(models[[i]], new_cases)
  }
  
  # Determine the majority vote
  final_predictions <- apply(do.call(cbind, predictions), 1, function(x) {
    names(sort(table(x), decreasing = TRUE))[1]
  })
  
  return(final_predictions)
}

Question 6

# Vectors to store results
accuracies <- numeric(5)
tprs <- numeric(5)
tnrs <- numeric(5)

# Pull metrics for each validation dataset
for (i in 1:5) {
  # Get the validation set
  val_set <- validation_data[[i]]
  
  # Make predictions
  predictions <- predict_party(nb_models, val_set)
  
  # Confusion matrix
  matrix <- table(Actual = val_set$Class, Predicted = predictions)
  
  # Calculate metrics
  accuracies[i] <- sum(diag(matrix)) / sum(matrix)
  tprs[i] <- matrix[2, 2] / sum(matrix[2,])
  tnrs[i] <- matrix[1, 1] / sum(matrix[1,])
}

# Mean performance across all folds
mean_performance <- c(
  Accuracy = mean(accuracies),
  TruePositiveRate = mean(tprs),
  TrueNegativeRate = mean(tnrs)
)

# Result table
result_table <- data.frame(
  Metric = names(mean_performance),
  Value = mean_performance
)
row.names(result_table) <- NULL
print(result_table)

##             Metric     Value
## 1         Accuracy 0.8928571
## 2 TruePositiveRate 0.9038510
## 3 TrueNegativeRate 0.8851244

Question 7

# Member voting record
new_member_votes <- data.frame(
  V1 = "n",
  V2 = "y",
  V3 = "n",
  V4 = "y",
  V5 = "y",
  V6 = "y",
  V7 = "n",
  V8 = "n",
  V9 = "n",
  V10 = "y",
  V11 = "n",
  V12 = "y",
  V13 = "y",
  V14 = "y",
  V15 = NA,
  V16 = "n",
  stringsAsFactors = FALSE
)

# Prediction using ensemble 
predicted_party <- predict_party(nb_models, new_member_votes)

# Add party name
party_name <- ifelse(predicted_party == 1, "democrat", 
                      ifelse(predicted_party == 2, "republican", NA))

print(party_name)

## [1] "republican"