Olumide | 1 Oct 01:50 2002
Picon

Re: problem in installation of weka

>
>
>
> I  tried to apply the instruction : java weka.classifiers.j48.J48
>-t weather.arff
>
>to test the j48 algo but i get this messge:
>
>Weka exception: Can't open file weather.arff (The system cannot find the
>file specified).
>

It looks like you are trying the command-line version of Weka. (The 
problem is that you need to tell Weka where to find the file 
weather.arff, so you will have to specify a path along with the filename 
e.g. D:\Work\weather.arff

I think its best you started learning Weka with the GUI versions. You 
can find an excellent tutorial at:
http://www.chem-eng.utoronto.ca/~datamining/Tutorials/Experiments.pdf. I 
learnt using it myself ...

Best Wishes

- Olumide
Timo Becker | 1 Oct 20:27 2002
Picon

StringToWordVector

Hi!

I have some problems converting Text data into a word vector with
weka.filters.unsupervised.attribute.StringToWordVector.
My example-file animals.arff contents

 <at> relation text

 <at> attribute docs string
 <at> attribute topic {animals, no_animals}

 <at> data

"cat mouse cat", animals
"tree flower", no_animals

When I convert this using StringToWordVector with the command
java weka.filters.unsupervised.attribute.StringToWordVector -i
animals.arff -o animals_out.arff
the file animals_out.arff contents

 <at> relation
'text-weka.filters.unsupervised.attribute.StringToWordVector-D.,:\'\"()?!-R1-w1000'

 <at> attribute topic {animals,no_animals}
 <at> attribute cat numeric
 <at> attribute flower numeric
 <at> attribute mouse numeric
 <at> attribute tree numeric

(Continue reading)

Len Trigg | 1 Oct 23:30 2002
Picon

Re: StringToWordVector

Timo Becker <beck2402 <at> ldv39.uni-trier.de> wrote:
> Why is the value for cat only 1 instead of 2?
> I expected {1 2,3 1} for the first instance.
> When I looked at the source-code of StringToWordVector I recognized that
> the word-frequencies are counted.

Word-frequency information is currently only used in determining
the set of words to use as attributes (you can tell the filter to have
a reduced dictionary for large datasets). 

> But somewhere they must have been lost.
> Does anybody know what to do?

I have added an option to output word counts to this filter. It is now
in CVS.

Cheers,
        Len.
M Al-yahya | 2 Oct 12:11 2002
Picon

CVS Files

Hello,
 Is it possible to access weka CVS files? If so, how can that be done?

Cheers,
mma
Timo Becker | 2 Oct 12:28 2002
Picon

Re: StringToWordVector

Hi!

I tried to login, but failed.

When I typed

export CVSROOT=:pserver:cvs_anon <at> cvs.scms.waikato.ac.nz:/usr/local/global-cvs/ml_cvs
cvs login
I was kicked after entering Return as password.

So i tried
cvs -d:pserver:cvs_anon <at> cvs.scms.waikato.ac.nz:2401/usr/local/global-cvs co weka
and got
/usr/local/global-cvs: no such repository

What did I do wrong?

Greetings,
Timo

Timo Becker LDV/Phonetik
beck2402 <at> uni-trier.de

On Wed, 2 Oct 2002, Len Trigg wrote:

> Timo Becker <beck2402 <at> ldv39.uni-trier.de> wrote:
> > Why is the value for cat only 1 instead of 2?
> > I expected {1 2,3 1} for the first instance.
> > When I looked at the source-code of StringToWordVector I recognized that
> > the word-frequencies are counted.
(Continue reading)

zied.elouedi | 2 Oct 19:28 2002
Picon

materiel ressources

Dear sir
I would apply weka for a huge databais of 1giga byte.

I would like to know with kind of materiel ressources should i use (at
least) to get results.

I have a pc 850mhz, 128 megabytes of memory
i use windows millinum
do you think is it sufficent as ressources
thanks a lot
regards

--

-- 
+++ GMX - Mail, Messaging & more  http://www.gmx.net +++
NEU: Mit GMX ins Internet. G√ľnstige DSL- & Modem/ISDN-Tarife!
Olumide | 3 Oct 06:10 2002
Picon

What meaneth these (SVM) numbers

Hi ...

My SVM generates these weird numbers following each classification.

[For example]
7 Big 0.31886580482427557
8 Big 0.3243092676529198
9 Small 0.389639428035945
10 Small 0.3554180378809114
11 Normal 0.3317574350700097
12 Big 0.3386343287150975

What do these numbers mean?

Thanks

- Olumide
Aldebaro Klautau | 3 Oct 18:20 2002
Picon

Re: What meaneth these (SVM) numbers

I assume this is the result obtained with option -p 0. I think the rule
is: if the classifier is a subclass of DistributionClassifier, the
"margin" is printed and there are 4 columns (instance number, correct
label, margin and recognized label). If it's a subclass of Classifier, the
margin is not printed and there are only 3 columns.

What confuses me in your listing below is that I don't see the fourth
column with the recognized labels...

 On Thu, 3 Oct 2002, Olumide wrote:

> Hi ...
>
> My SVM generates these weird numbers following each classification.
>
> [For example]
> 7 Big 0.31886580482427557
> 8 Big 0.3243092676529198
> 9 Small 0.389639428035945
> 10 Small 0.3554180378809114
> 11 Normal 0.3317574350700097
> 12 Big 0.3386343287150975
>
> What do these numbers mean?
>
> Thanks
>
>
> - Olumide
>
(Continue reading)

Len Trigg | 4 Oct 04:16 2002
Picon

Re: StringToWordVector

Timo Becker wrote:
> export CVSROOT=:pserver:cvs_anon <at> cvs.scms.waikato.ac.nz:/usr/local/global-cvs/ml_cvs
> cvs login
> I was kicked after entering Return as password.

The cvs login command normally returns to your command prompt
afterwards, so this looks fine.

> cvs -d:pserver:cvs_anon <at> cvs.scms.waikato.ac.nz:2401/usr/local/global-cvs co weka
> and got
> /usr/local/global-cvs: no such repository

This command should be:

cvs -d:pserver:cvs_anon <at> cvs.scms.waikato.ac.nz:/usr/local/global-cvs/ml_cvs co weka

(you were missing the ml_cvs)

Cheers,
        Len.
Len Trigg | 4 Oct 04:33 2002
Picon

Searching the list archives...


For those wanting to search the list posts (there were some posts a
while back about this), you can accomplish this with google by adding
site:list.scms.waikato.ac.nz to your google query, e.g.:

http://www.google.com/search?q=site%3Alist.scms.waikato.ac.nz%20weka%20neural

Cheers,
        Len.

Gmane