draft/na_filter.R

In filipezabala/voice: Voice Analysis, Speaker Recognition and Mood Inference via music theory

#'  NA filter
#'
#' @description Filters
#' @param x An object of class \code{aggr} from \code{VIM} package.
#' @param max.loss A directory/folder containing RTTM files.
#' @examples
#' \dontrun{
#' library(voice)
#' # http://archive.ics.uci.edu/ml/datasets/Arrhythmia
#' dat <- 'http://archive.ics.uci.edu/ml/machine-learning-databases/00579/MI.data'
#' bd <- read.table(dat, sep = ',')
#' bd[bd=='?'] <- NA
#' na_filter(bd)
#' }
#' @export
na_filter <- function(x, max.loss = 1){
  d <- x$missings
  d.ord <- d[with(d, order(-Count, Variable)), ]
  d.ord$Prop <- d.ord[,'Count']/nrow(x$x)
  filtro <- d.ord[d.ord$Prop <= max.loss, 'Variable']
  return(filtro)
}
#
# # exploratória
# miss = VIM::aggr(bd, sortVars = T)
# mm = miss$missings
# quantile(round(mm$Count/nrow(bd),3), probs = seq(0,1,.1))
#
# # função para listar as variáveis com perda máxima limitada (1 traz tudo)
#
# na_filter(miss) # tudo, padrão
# na_filter(miss, .05) # perda máxima (NA) de 5%
# na_filter(miss, .005) # perda máxima (NA) de 0.5%
#
# # filtrando as variáveis com no máximo 0.5% de perda
# library(tidyverse)
# bd2 <- bd %>%
#   dplyr::select(sort(na_filter(miss, .005)))
# dplyr::glimpse(bd2)
#
# VIM::aggr(bd2, sortVars = T)