Orange Forum • View topic - Help predicting please

Help predicting please

A place to ask questions about methods in Orange and how they are used and other general support.

Help predicting please

Postby Guest » Tue Dec 08, 2009 15:06

Here is the situation I have:

I have a data set that looks like this, about 20,000 rows in this format:

Code: Select all
X,Y,A,B,C


where X and Y won't change (information about people) and A, B, and C are true/false values for that person over the past 3 years.

What I have done so far is use a 10% random sample of this to train the logistic and classification tree models, then connect these to the learner tester with the full data source as its data input. Both classifiers perform very well and are able to model the full class C based on X, Y, A, and B.

I also have a data set that looks like this:

Code: Select all
X,Y,B,C


that represents the next year's worth of data (each data set looks at the past 3 years in the columns: A-C this year and B-D is what it would be next year). The problem is that I do not know the outcome of any of the "D" values yet (outcomes next year based on X, Y, B and C, C being what was modeled for this year based on known outcomes). I want to use my training model to predict D but I'm not sure how. Some help please? I have been using the visual interface so far but if I need to do it via Python scripting that's fine.

Also, how can I make the logistic regression reveal the mathematical formulas it has derived? I can't see any way to see this so I can't simply apply its model in software outside of Orange at present.

Thank you!

Return to Questions & Support