To Extend Or Include?

[tetsu yatsu]

This is a basic question, and I'm mostly testing to see if this mailing list is still alive.

 

 

I've been trying to write a Weka application for a few days now. I've gotten used to the GUI, and I see that the Java end of it is very similar. I have two potentially basic questions:

 

I have my own CSV data formats. I have trained model files to sufficient accuracy. I want to predict what the missing values are and replace the missing CSV values with the predicted value, but to do the prediction I need to Remove the extraneous attributes that would otherwise muck up the prediction. Should I create a new class for my CSV data format and create a Weka CSV loader object as a member, or should I extend the CSV loader class to include my whole CSV file?

 

And the second question; I have to run a bunch of filters. Is there a way to chain many filters together? Currently I'm resetting a dataset variable to the outcome of the next filter, hoping garbage collection will handle the memory allocation for me.

 

 

Hope you guys are still around. I like your software package, and look forward to using it in many future projects!

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Peter Reutemann-3]

> This is a basic question, and I'm mostly testing to see if this mailing list
> is still alive.

Huh? Why shouldn't it be alive? Just check the Wekalist archives. That
should answer your question...

> I've been trying to write a Weka application for a few days now. I've gotten
> used to the GUI, and I see that the Java end of it is very similar. I have
> two potentially basic questions:
>
> I have my own CSV data formats. I have trained model files to sufficient
> accuracy. I want to predict what the missing values are and replace the
> missing CSV values with the predicted value, but to do the prediction I need
> to Remove the extraneous attributes that would otherwise muck up the
> prediction. Should I create a new class for my CSV data format and create a
> Weka CSV loader object as a member, or should I extend the CSV loader class
> to include my whole CSV file?

Why don't you just use the FilteredClassifier meta-classifier with
your base-classifier of choice and the Remove filter to remove the
unwanted attributes? Then you can just load the full data and run it
through the meta-classifier without having to worry about applying the
filter separately.

> And the second question; I have to run a bunch of filters. Is there a way to
> chain many filters together? Currently I'm resetting a dataset variable to
> the outcome of the next filter, hoping garbage collection will handle the
> memory allocation for me.

Use weka.filters.MultiFilter to apply all of the filters in one go.

> Hope you guys are still around. I like your software package, and look
> forward to using it in many future projects!

Please keep in mind: if you're using Weka in a project that is
publicly available, all of your code has to be released under the GPL
and made available for download as well. Otherwise, you'll have to
obtain a commercial license from the university (through Waikatolink).

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html