MEKA GUI - Performance

[LeguDave]

Hello,

I recently started using MEKA GUI and I have a data set similar to your example music.arff. However the 10 fold cross validation of my data set does not terminate or takes longer than 3 hours. Although a default train/split takes roughly half a minute.

My data set has

• 3600 instances
• about 1500 numeric attributes
• Goal: classify the first three attributes (labels)

Configuration

• Binary Relevance
• J48 -C 0.25 -M 2
• Cross Validation 10 Fold

My questions are:

• Is this an expected computing time?
• If yes, is it because I have that many attributes?
• What can I do to increase performance?

[SOLVED]

Used the same configurations again later and it worked. Maybe reopening the MEKA GUI after changing data sets solved this issue.

[fracpete]

It would be good know whether you can replicate this behavior as it might point to a bug in MEKA.