CODEWITHSUNDEEP
r
dasur schn
dasur schn

Reputation: 61

NAs introduced by coercionError in randomForest.default(m, y, ...) : NA/NaN/Inf in foreign function call (arg 1)

I am running following code

class(TrainSet$volume) that gives me [1] "numeric"

Then I run

model1 <- randomForest(TrainSet$volume ~ ., data = TrainSet, importance = TRUE)

it gave me

Error in randomForest.default(m, y, ...) : NA/NaN/Inf in foreign function call (arg 1)

What could be the reason? Thanks

Upvotes: 1

Views: 623

Answers (2)

Michael McFarlane
Michael McFarlane

Reputation: 131

Besides missing data or infinities as Peter_Evan suggested, another possibility is a character variable in TrainSet. Therefore, run the following four lines:

any(is.na(TrainSet))
any(apply(TrainSet, 2, is.infinite))
any(apply(TrainSet, 2, is.nan))
any(is.character(TrainSet))

If any return TRUE, you have your problem.

Upvotes: 1

Peter_Evan
Peter_Evan

Reputation: 947

It is hard to know for sure without more information about your data, but as the error suggests you seem to have one of those values (NA/NaN/Inf) somewhere in your data frame. Perhaps inf as NA tends to throw a different error. We can recreate your error below:

library(randomForest)

#setting data
data(iris)

#making an infinite value
iris[1,1] <- Inf

#grab row
iris[is.infinite(iris$Sepal.Length),]

#output
#   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
# 1          Inf         3.5          1.4         0.2  setosa

#checking data type
is.numeric(iris$Sepal.Length) #TRUE

#reproducing error
iris.rf <- randomForest(iris$Sepal.Width ~ ., data=iris, importance=TRUE)

#output
Error in randomForest.default(m, y, ...) : 
  NA/NaN/Inf in foreign function call (arg 1)

As to where or why this is in your data is unclear (again, need to see the data to make this call). A common way inf is created is by a mistake in per-processing that introduces a confused calculation, like dividing be zero. 

is.infinite(pi / 0)
#output
# [1] TRUE

Scanning for infinites or NA (with is.infinite or is.na) and reviewing any changes you made to your data seems like a good place to start.

Upvotes: 2

Related Questions