Incorrect Learner Properties and Feature Types in mlr3shiny

[karo588]

trafficstars

All learners will state that they support the “missings” property and can handle all data types, although this is not always the case. For example, SVM cannot handle missing values and only supports the feature types logical, integer and numeric. However, mlr3shiny states that SVM supports the “missings” property, as well as the feature types “character”, “factor”, “ordered” and “POSIXct”.

Steps for reproducibility:

1. Add the print(learnerobject$Learner) command in the server_learner.R script after the line learnerobject$Hash <- learnerobject$Learner$hash of the makeLearner() function.
2. Start mlr3shiny.
3. Select the “german_credit” task, for example.
4. Select learner “Support Vector Machine” and click on “Go”.
5. Console displays the wrong Properties and Feature Types.