R/my_knn_cv.R

In SylviaDu99/Project3Package: Package Building Project

#' k-nearest neighbors function
#'
#' This function predicts the class value by k-fold Cross-validation and k-nearest neighbors algorithms.
#'
#' @param train Dataframe of training data
#' @param cl Vector input of true class value of the training data
#' @param k_nn Numeric number of neighbors used in k-nearest neighbors
#' @param k_cv Numeric number of folds used in k-fold cross-validation
#' @keywords prediction
#'
#' @return List object containing the predicted class value of the data, the CV
#' misclassification rate and the training misclassification rate
#'
#' @examples
#' my_knn_cv(my_gapminder[, 4:5], my_gapminder$continent, 1, 5)
#'
#' @import class stats magrittr tidyverse dplyr
#'
#' @export

my_knn_cv <- function(train, cl, k_nn, k_cv) {
    # check if k_nn and k_cv are both numeric, if not, stop and show error information
    if (!is.numeric(k_nn) || !is.numeric(k_cv)) {
        stop("k_nn and k_cv must be numeric")
    }
    # split data in k_cv parts, randomly
    fold <- sample(rep(1:k_cv, length = length(cl)))
    data <- data.frame("x" = train, "y" = cl, "split" = fold)
    class <- c()
    cv_err <- rep(NA, k_cv)
    fold_l <- length(cl) / k_cv
    # iterate through i = 1 to k_cv
    for (i in 1:k_cv) {
        # predict class value of the ith fold using all other folds as training data
        data_train <- data %>% dplyr::filter(split != i)
        data_test <- data %>% dplyr::filter(split == i)
        y_hat <- knn(select(data_train, contains("x")), select(data_test, contains("x")), data_train$y, k_nn)
        class <- c(class, y_hat)
        # record the prediction and the misclassification rate
        cv_err[i] = sum(y_hat != data_test$y) / fold_l
    }
    y_hat <- knn(train, train, cl, k_nn)
    train_err = sum(as.numeric(y_hat != cl)) / length(cl)
    output <- list("class" = class, "cv_err" = mean(cv_err), "te" = train_err)
    return(output)
}