Forum Stats

  • 3,760,198 Users
  • 2,251,663 Discussions
  • 7,871,021 Comments

Discussions

Applying a Model and the Columns Generated

1006939
1006939 Member Posts: 1
edited May 14, 2013 8:11PM in Machine Learning
Hi,

I wanted some help regarding the output of a decision tree model, i am building a model that predicts churners, the dataset consists of 20% churners depicted as 'Y'. The model is built and when i apply the model it seems to give higher %'s to 'N' rather than 'Y'. I was expecting to see the % of 'Y' higher. Upon closer inspection i saw that the DT Viewer itself was solving for 'N'. Is there anyway to change this default behaviour.

2nd QUestion

In the Apply table, if there are two rows one is prediction 'Y' and prob .94 the other is prediction 'N; and prob .89, do i read this as there is a 94% chance he will be a Y and a 89% chance he will be a N?

Please help

Regards
Sam
Tagged:

Answers

  • Hi Sam,

    The Apply functions allow you to choose what target class value you want to predict.
    The default generated prediction function is most likely.
    If you prefer to just get N or Y then you can choose the specific class value prediction via the Apply Node editor (ODMr terminology) which then invokes the data mining sql scoring functions.

    I think you are interpreting the apply results correctly.
    See the following link for more details on the ODM sql scoring functions:
    http://docs.oracle.com/cd/E11882_01/server.112/e26088/functions136.htm#SQLRF06212

    Thanks, Mark
This discussion has been closed.