ARFFVectorIterable/MapBackedArffModel doesn't handle question mark '?', other
ARFF issues
-----------------------------------------------------------------------------------------
Key: MAHOUT-952
URL: https://issues.apache.org/jira/browse/MAHOUT-952
Project: Mahout
Issue Type: Bug
Components: Integration
Affects Versions: 0.6
Environment: Latest SVN on ubuntu
Reporter: Stuart Smith
Priority: Minor
Whatever is parsing the ARFF file for the ARFFVectorIterable (As far as I can
tell, it's the class itself) doesn't handle '?' as a marker for unknown value.
See: http://www.cs.waikato.ac.nz/~ml/weka/arff.html
I just started looking at Mahout classifiers this week, so I'm not sure how to
handle this yet. If I figure it out, I'll post a patch, but until then,
guidance would be helpful!
Off topic, but I'm also having some issue were the labels populated in the map
apparently aren't coming from the Attribute Header at the top of the file. I
have very sparse vectors (1800+ attributes, only a few hundred set for any
given before).. and I keep getting IndexOutOfBounds or mismatched cardinality
issues, depending on whether I use full ARFF or sparse ARFF. Either way, when I
dump the Labels from getModel(), it doesn't have them all.. even if I parse the
ARFF myself, and call setLabel() (Apparently just throws that away). Looks like
the DenseVectors keep thinking the cardinality is 534, when it should be
1800+.... when I know more, I'll create a new issue
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
