Orange Forum • View topic - Saving a model and testing against new data set

Saving a model and testing against new data set

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

Saving a model and testing against new data set

Postby nickh82 » Mon May 06, 2013 15:15

Hi there,

I'm trying to get my head around how to use this tool from visual canvas. I have no experience in Python so cannot use that method. I just wanted to know two things:

1. Are there any good sources of documentation on the use of canvas orange? Any Youtube portals or lecture presentations etc.
2. I have build a k-means clustering model and am trying to use that model to test against another data set to see how well trained it is. But I have no idea how to use the model as an object to apply it to a second set of data. Can anyone help?

Thanks for your time.

Re: Saving a model and testing against new data set

Postby Ales » Wed May 08, 2013 12:43

nickh82 wrote:1. Are there any good sources of documentation on the use of canvas orange? Any Youtube portals or lecture presentations etc.
There are some tutorials/presentations on the web (for instance http://wiki.sdakak.com/ml:getting-started-with-orange, also google is your friend)

nickh82 wrote:2. I have build a k-means clustering model and am trying to use that model to test against another data set to see how well trained it is. But I have no idea how to use the model as an object to apply it to a second set of data.
You can not use the k-means directly to apply to a data set.
You can however train a k nearest neighbor classifier on the k-Means "Centroids" output table and use it to 'predict' the induced clusters on unseen data. You just have to make sure the selected 'Distance' measure in both the k-Means and k-Nearest Neighbours widget is the same.

Then you can use the 'Predictions' widget to predict and observe the clusters on new data.


Return to Questions & Support



cron