The Learners (SVM, Decision Tree, KNN) all default to using the Document Class as the predictor (without ability to change it). How can I set this value upstream in the node path? Is there another way to change the predictor column in these learners? Some of the Predictors allow for this. Note that I have tried every conceivable permutation of using Strings to Documents node to map Document Category to the Document Class (which would solve the problem) but none of the permutations work. I always get an error that input values are not strings (again, I have played with many options to include Meta Info Extractor Inserter / Extractor, Document Data Extractor, etc. to convert the input from Document, Document Cell, Table to strings). Any help would be very much appreciated.
The Strings to Document node allows to set the values of a string column as document category or source. This field can be extracted later on again as string and be used as e.g. class / target column for learner nodes. If you have a numerical value that you want to set as category you need to bin it or convert it into a string (number to string). Only the Strings to Document node allows to set the category and source fields. Alternatively you can use the Meta Inserter and Extractor nodes to insert meta information as key value pairs.
If you need more specific advise how to use these nodes please share a workflow with data and I will try to set up an example for you.
Cheers, Kilian
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.