R/algos.R

In cchoe/numerator: R Functions commonly used for Data Scientists at Numerator

#' K-means optimal cluster calculation using elbow method
#'
#' Low-stats NbClust instability solved for data frames with less than 5 rows
#'
#' @param x x coordinate
#' @param y y coordinate
#' @param k.max max k clusters that can be generated (Default=6).
#' @return kmeans cluster object, containing means, clustering vector, etc.
#' @examples kmeans_optimal_elbow(x_coord, y_coord)
#' @export
kmeans_optimal_elbow <- function(x, y, k.max=6) {
  df1 = as.matrix(cbind(x, y))
  n_clusters = nrow(unique(df1))
  if (n_clusters > 1) {
    if (nrow(unique(df1)) < 5) {
      a1 = fviz_nbclust(df1, kmeans,
                        k.max = k.max,
                        method = "wss",
                        print.summary = T,
                        verbose = T)
      plot(a1)
      q = a1$data$y/a1$data$y[1]
      n_clusters = ifelse(min(which(q < 0.15)) == Inf &
                              nrow(unique(df1)) == 3, 2,
                          min(which(q < 0.15)))
      clusters = kmeans(df1 ,n_clusters)
    } else {
      nb = NbClust(data = df1,
                   diss = NULL,
                   distance = "euclidean",
                   min.nc = 2,
                   max.nc = k.max,
                   method = "kmeans",
                   alphaBeale = 0)
      n_clusters = max(unlist(nb[4]))
      print(n_clusters)
      clusters = kmeans(df1, n_clusters)
    }
  }

  return(clusters)
}