Orange Forum • View topic - Sparse data file

Sparse data file

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

Sparse data file

Postby Viktor » Fri Jul 21, 2006 16:15

Hi,

It would be really nice to be able to input sparse data representations. There is the basket format, but it seems it can be used only for association rules. The documentation on basket says:

"Examples in such format can be used for derivation of association rules; if you want to use them for any other purpose, you need to convert them into "ordinary" examples by pulling some meta-attributes into ordinary attributes."


But it is not very clear what "pulling some meta-attributes into ordinary attributes" means. Any help will be appreciated.

Postby Janez » Fri Jul 21, 2006 23:35

Sparse data sets may have thousands of attributes, which are here stored as meta-attributes. To convert the data to non-sparse data, you would select, say, ten attributes, construct a new Domain and convert the table. The used meta attributes will be copied and the non-used will be discarded.

But this is probably not what you wanted. The documentation is not really accurate here: of course you can use such sparse data tables, but no (existing) learning algorithm is able to learn from meta attributes. You cannot construct a classification tree from sparse data, but if you'd like to try an algorithm you program yourself, you can do it.

Honestly, Orange's support for sparse data and for text mining is rather basic, but we work on that. (Contributions are welcome.)

Postby vim » Sun Jul 23, 2006 3:16

Janez wrote:Honestly, Orange's support for sparse data and for text mining is rather basic, but we work on that. (Contributions are welcome.)

good news to me, can we get more information about text mining/classification with orange?

Postby Blaz » Fri Sep 08, 2006 18:40

The relevant topic for this forum is just active, see Too many Attributes.


Return to Questions & Support