encode_cat.R

# create dummy data for testing
require(caret)
require(dplyr)
full <- data.frame(target = sample(c(0,1), 500, replace=T),
                   ID = 1:500,
                   v1 = sample(LETTERS[1:2], 500, replace=T),
                   v2 = sample(1:100, 500, replace=T),
                   v3 = sample(LETTERS[1:10], 500, replace=T),
                   stringsAsFactors = FALSE)
folds <- createFolds(full$target, k=5, list=TRUE, returnTrain=FALSE)
train <- full

#--------------------------------------
## original code from happy
ohe.list <- names(full[, sapply(full, is.character)])
# > ohe.list
# [1] "v1"   "v3"
list.x <- list()
for (i in seq(1, length(folds)) ) {
  lofo.ids <- folds[[i]]
  x <- train[-lofo.ids, ] %>%
    group_by(v1) %>%
    mutate(v1_avg = mean(target, na.rm = T)) %>%
    ungroup %>%
    select(ID, v1_avg)
  list.x[[i]] <- x
}
list.x <- do.call(rbind, list.x)
# Source: local data frame [2,000 x 2]
# 
# ID    v1_avg
# (int)     (dbl)
# 1      1 0.5242718
# 2      2 0.4639175
# 3      3 0.4639175
# 4      6 0.4639175
# 5      7 0.5242718
# 6      8 0.4639175
# 7      9 0.5242718
# 8     10 0.4639175
# 9     11 0.5242718
# 10    12 0.4639175
# ..   ...       ...

#--------------------------------------
## revised code 

encode_cat <- function(mydat) {  # col1=target, col2=categorical var
  names(mydat)[2] <- "myvar"
  target.avg <- mydat %>% group_by(myvar) %>%
    mutate(v_avg = mean(target, na.rm = T)) %>%
    ungroup %>% select(v_avg)
  return(target.avg)
}

ohe.list <- names(full[, sapply(full, is.character)])  # v1, v3
list.x <- list()
for (i in seq(1, length(folds)) ) {
  lofo.ids <- folds[[i]]
  train.foldi <- train[-lofo.ids, ]
  x <- train.foldi %>% select(ID)
  for (j in 1:length(ohe.list)) {
    x <- bind_cols(x, encode_cat(train.foldi[c("target", ohe.list[j])]))
    names(x)[j+1] <- paste0(ohe.list[j],"_avg")
  }
  list.x[[i]] <- x
}
list.x <- do.call(rbind, list.x)
# Source: local data frame [2,000 x 3]
# 
# ID    v1_avg    v3_avg
# (int)     (dbl)     (dbl)
# 1      1 0.5242718 0.5151515
# 2      2 0.4639175 0.4615385
# 3      3 0.4639175 0.5151515
# 4      6 0.4639175 0.5000000
# 5      7 0.5242718 0.4705882
# 6      8 0.4639175 0.4047619
# 7      9 0.5242718 0.4705882
# 8     10 0.4639175 0.5833333
# 9     11 0.5242718 0.4615385
# 10    12 0.4639175 0.5625000
# ..   ...       ...       ...