Error in randomforest.default(m, y, ...) : na/nan/inf in foreign function call (arg 1)

Question

Problem :

I have done a lot of research on this extensively without finding any solution on it. I have tried cleaning my data set as follows:

library("myraster")

impute.mean <- function(l) replace(l, is.na(l) | is.nan(l) | is.infinite(l) , 

mean(l, na.rm = TRUE))

losses <- apply(losses, 2, impute.mean)

colSums(is.na(losses))

isinf <- function(l) (NA <- is.infinite(l))

infout <- apply(losses, 2, is.infinite)

colSums(infout)

isnan <- function(l) (NA <- is.nan(l))

nanout <- apply(losses, 2, is.nan)

colSums(nanout)

But the problem arises while running the predict algorithm:

options(warn=2)

p  <-   predict(default.rf, losses, type="prob", inf.rm = TRUE, na.rm=TRUE, nan.rm=TRUE)

All my research says it should be NA's or Inf's or NaN's in the data but I don't have any

Answer 1

I guess you are failing to replace Inf values with the means of column vectors. This is happening because of the na.rm = TRUE argument in your call to mean() within the impute.mean function which does exactly what it says i.e. it removes NA values and clearly not Inf ones

You can cross check it, by following code example:

impute.mean <- function(l) replace(l, is.na(l) | is.nan(l) | is.infinite(l), mean(l, na.rm = TRUE))

losses <- apply(losses, 2, impute.mean)

sum( apply( losses, 2, function(.) sum(is.infinite(.))) )

# [1] 696

To remove the infinite values, please use following:

impute.mean <- function(l) replace(l, is.na(l) | is.nan(l) | is.infinite(l), mean(l[!is.na(l) & !is.nan(l) & !is.infinite(l)]))

losses <- apply(losses, 2, impute.mean)

sum(apply( losses, 2, function(.) sum(is.infinite(.)) ))

# [1] 0