Open CarolynOlsen opened 6 years ago
This looks like a limitation in the C5.0 C code. You can escape other characters but I've been testing a bit and it doesn't accept this inside the data values.
You might dummy up some application files to verify. If it doesn't work, I'd email RuleQuest and see if Quinlan can make a change.
Same problem here: had no problem before, but after upgrading, commas in variables break the training proccess :-(
Will check escaping the commas and report back...
(EDITED):
Sorry, don't have time... I've downgraded with install_version("C50", version = "0.1.0-24", repos = "http://cran.us.r-project.org")
to get the old comma-tolerant functionality...
C5.0() now fails on factor variables that include commas, where it did not before.
I recently updated my version of C50, and tried to train a model on a data set I've trained C5.0 models on before. I now receive the error "c50 code called exit with value 1". I narrowed it down to one factor variable that had commas in the values. After removing the commas, the model trained fine. Below is a small example I created to replicate the problem.
Thank you very much!