"Force neural net output value"

navnav MemberPosts:28Maven
edited May 2019 inHelp
Hello, is it possible to force the neural net output value to positive value? My problem is that with a training set with positive and negative attributes but an only positive label attribute, the neural net classifier sometimes predicts wrong negative values. Thanks a lot to all.
Tagged:

Answers

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, MemberPosts:2,531Unicorn
    Hi,
    you could simply apply a postprocessing step. You could easily configure the Generate Attribute operator to derive a column that has a minimum of 0 for example. Then you could set the role of this new column to "prediction" and it will be used as new prediction in a following performance estimator.


    Greetings,
    Sebastian
Sign InorRegisterto comment.