R - convert dataframe columns to numeric - NAs introduced by coercion error

Question

I am trying to execute the code from here:

Change the class from factor to numeric of many columns in a data frame

in a dataframe with 140 columns

cols = c(1:140);    
merged_dataset[,cols] = apply(merged_dataset[,cols], 2, function(x) as.numeric(as.character(x)));

the problem is for some columns I get NAs. Is there a way somehow exclude these columns from the code so that I keep the data and they don't get transformed into NAs? I see the type of these columns is character if that helps.

Answer 1

If you already know the indices of the columns you want to drop, then you may subset your data frame to target only certain columns:

cols <- c(1:140)         # all columns
cols.skip <- c(1,3,5,21) # columns which CAN'T be converted to numeric
cols.keep <- cols[!cols %in% cols.skip]
merged_dataset[,cols.keep] <- apply(merged_dataset[,cols.keep], 2, function(x) {
    as.numeric(as.character(x))
})

To implement similar logic using column names rather than indices:

cols.skip <- c("a", "b", "c")
cols.keep <- !(names(merged_dataset) %in% cols.skip)
merged_dataset[,cols.keep] <- apply(merged_dataset[,cols.keep], 2, function(x) {
    as.numeric(as.character(x))
})

Answer 2

Substitution of any improper characters inside factor levels can also occur to better extract any numbers:

convert_factors_to_numeric <- function(df) {

  as.data.frame(lapply(df,
                       function(x) {

                         if (is.factor(x)) {

                           as.numeric(as.character(trimws(x),
                                        which = "both"))

                         } else{

                           x

                         }
                       }
                     ),

                stringsAsFactors = FALSE)

}

df_converted <- convert_factors_to_numeric(df)