"No covariates found" problem with matrix input

[nikolausthomsen]

Hi, many thanks for the ranger package! Is there a way to directly use a matrix as a covariate without colnames? In this version of the package I don't see the solution...

I hope the code explains the issue well. x is a given matrix but without colnames ranger does not accept it.
`

x <- matrix(rnorm(30*2), ncol=2); y <- rnorm(30)
ranger(x=x,y=y)
Error in ranger(x = x, y = y) : Error: No covariates found.
colnames(x)
NULL
colnames(x) <- paste0("x",1:2)
ranger(x=x,y=y)
Ranger result

Call:
ranger(x = x, y = y)

Type: Regression
Number of trees: 500
Sample size: 30
Number of independent variables: 2
Mtry: 1
Target node size: 5
Variable importance mode: none
Splitrule: variance
OOB prediction error (MSE): 1.034131
R squared (OOB): -0.4679862

`

[mnwright]

Currently, we are relying on the column names to re-identify the correct features during prediction. In principle, we could accept matrices without colnames and just expect the same feature order for prediction.

[SamGG]

Thanks for your feedback.

IMHO, ranger would benefit for setting colnames when it's NULL.

As the code of the OP shows it, when no name is provided to columns, there is an error and the code stops.

x <- matrix(rnorm(30*2), ncol=2); y <- rnorm(30)
ranger(x=x,y=y)
Error in ranger(x = x, y = y) : Error: No covariates found.

So, there is no result at all and I don't understand

[we] just expect the same feature order for prediction.