How to build random forests in R with missing (NA) values?

Question

I would like to fit a random forest model, but when I call

library(randomForest)

cars$speed[1] <- NA # to simulate missing value

model <- randomForest(speed ~., data=cars)

I get the following error

Error in na.fail.default(list(speed = c(NA, 4, 7, 7, 8, 9, 10, 10, 10,  : missing values in object

Answer 1

If you want to train a random forest with data having missing values, then maybe you have some misconceptions about training a model. 

Training a machine learning model means, we are giving input data to model so that a specific algorithm can draw a pattern from that pattern. If you train any machine learning model with missing data, then maybe you are doing it wrong.

You should pre-process your dataset if you’re having missing value. You can use an imputer to fill missing values. 

If you are having a really small data set (that’s not useful for machine learning), then you can set this parameter like this:

na.action = na.omit

Hope this answer helps you a bit.