RegressionDemo id2outcome.txt missing some predictions and actual results

[GoogleCodeExporter]

id2outcome.txt in RegressionDemo is missing some predictions and actual 
results.  It looks like this:
STS.input.MSRvid.txt-737=3.400053;2.5
STS.input.MSRvid.txt-736=3.399867;2.75
STS.input.MSRvid.txt-735=;
STS.input.MSRvid.txt-734=;
STS.input.MSRvid.txt-733=;
STS.input.MSRvid.txt-732=3.400053;3.5

It should look like this: (my artificial numbers)
STS.input.MSRvid.txt-737=3.400053;2.5
STS.input.MSRvid.txt-736=3.399867;2.75
STS.input.MSRvid.txt-735=3.40016;2.5
STS.input.MSRvid.txt-734=3.40016;4.25
STS.input.MSRvid.txt-733=3.40016;3.2
STS.input.MSRvid.txt-732=3.400053;3.5

output/predictions.arff is not missing any predictions or actual results.

To replicate, run RegressionDemo, and look at 
WekaTestTask-Regression[...]/id2outcome.txt.

Original issue reported on code.google.com by EmilyKJa...@gmail.com on 11 Sep 2014 at 12:39

[GoogleCodeExporter]

The report tries to work with numeric outcomes by catching exceptions, however 
in regression mode, the outcome is also numeric, but if the gold label is <1 
the exception is not raised, but an empty string is returned, thus the empty 
lines.

I guess we need a special branch for regression ...

Original comment by torsten....@gmail.com on 11 Sep 2014 at 3:55

[GoogleCodeExporter]

Sorry for reporting late.
I'm working on this, as Torsten said, the problem occurs for predictions <1. 
This can easily be fixed bu there are other problems which made me turn 
WekaAdapter into WekaRegressionAdapter and WekaClassificationAdapter. I'm 
running a couple of tests and will commit soon.  

Original comment by daxenber...@gmail.com on 11 Sep 2014 at 4:01

• Changed state: Started

[GoogleCodeExporter]

This issue was closed by revision r1098.

Original comment by daxenber...@gmail.com on 11 Sep 2014 at 4:27

• Changed state: Fixed