Rapidminer: Filter Missing Labels does not work properly

Today, Rapidminer drove me nuts.

For my master thesis, I do a lot of text classification. However, I input a CSV file with examples having a class and having no class. Obviously, the ones with classes should be used for training and the rest for prediction. To reach this, splitting of the data is necessary, reasonable by the existence of a class (Or in the jargon of rapidminer: if a label is missing or if it is not missing).

However, it did not work. I've created a filters with the condition classes 'missing_labels' and 'no_missing_labels'. Nor of them did the work, they either gave me all examples or none. Just for completeness: Only one attribute was declared as label, I also tried to set the role explicitely by the "Set role" operator. This did not change the behaviour. 


I finally created a condition class 'custom_filters' and check for the question mark.

Leave a Comment

comments powered by Disqus