Peter Reutemann | 1 Mar 03:00 2009
Picon
Picon

Re: Connecting to MySQL database problem

> I want to connect Weka to my MySQL database to get some data. I've followed
> the instructions on http://weka.wiki.sourceforge.net/Databases and
> http://weka.wiki.sourceforge.net/CLASSPATH I've also check that weka has
> read that classpath: java.class.path:
> ./weka.jar;C:/Users/Krystal/Documents/University/University/Year 3/Data
> mining/Weka-3-6/mysql-connector-java-3.1.8-bin.jar

The "java.class.path" property is only a string. Java does not check
whether the entries listed there are actually available or valid. Only
at runtime, when a class is being instantiated, this classpath is used
to look for that class. Make sure that you don't have a typo in there.
If you've made sure that there is not typo, but still get the error
message that there's no drivers, you might want to try putting the
mysql jar in a location without blanks, e.g.,
"C:\jars\mysql-connector-java-3.1.8-bin.jar" and try again.

[...]

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

(Continue reading)

Peter Reutemann | 1 Mar 03:03 2009
Picon
Picon

Re: Weka graphs from ARFF file

> I've this from Internet and forget the source.

Here's the original wiki article for the source code:
  http://weka.wiki.sourceforge.net/Generating+ROC+curve

> See the code:
>
> public static void genereteROC(Evaluation eval) throws Exception {

[...]

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

Peter Reutemann | 1 Mar 03:06 2009
Picon
Picon

Re: How to get classification result

> How
> to know the classification of each instance. For example, for the
> iris, how to get the classification for each data given by weka.

Read FAQ "How do I make predictions with a trained model?". Link to
FAQs available from the Weka homepage.

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

Peter Reutemann | 1 Mar 03:14 2009
Picon
Picon

Re: Error:Nominal value not declared......WHY????

> I solved one problem i had but now i dont know why Weka says that i have
> this mistake:
>
> nominal value not declared in header, read token[_HOYERFELD_4],line 123
>
> My file is here:
>
> http://rapidshare.com/files/202027683/try2.arff.html

Check line 123. Your dataset defines 6 attributes, but this line has
the following 10 values:

16001097-2340-551224-001
SALLIX
October
82283.24
WILLIAM_KEREKNERA
_HOYERFELD_4
_33243_BIELD
_GERMANY
_INV.2858
PEKAO_BANK_O/KATOWICE

If a value contains a comma, you need to quote the value with double
or single quotes. You can use the method "quote" of the
weka.core.Utils class for this purpose.

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
(Continue reading)

Peter Reutemann | 1 Mar 04:17 2009
Picon
Picon

Re: Problems with remote engine setup

>  I'm trying to run a series of remote experiments, and I'm having
> trouble getting the system to work.  I hope someone here can help me
> out.
>
> What I'm trying to do:
> I have accounts on a number of remote machines running some flavor of
> unix, and a single Windows box on my desk.  I'd like to run a gamut of
> experiments using the remote machines as Remote Engines, and using the
> Windows box to run the Experimenter.
>
> The Remote Engines seem to be running without complaint.  They start
> up fine with the Java command line given in the manual, and spit out
> some messages along the lines of "RemoteEngine bound in RMI registry",
> and about once every ten minutes they'll report "purge -- no tasks to
> check".
>
>
> I have the HSQLDB server running on the Windows box, and it appears to
> have started up successfully with an incantation like the following:
>
> java -classpath "C:\Documents and
> Settings\myusername\Desktop\mydirectory\hsqldb.jar" org.hsqldb.Server
> -database.O experiment -dbname.O experiment
>
> I modified DatabaseUtils.props to include the line,
>
> jdbcURL=jdbc:hsqldb:hsql://127.0.0.1/experiment
>
> ...And since my database is running locally, I dropped
> DatabaseUtils.props into a directory along with hsqldb.jar and
(Continue reading)

Peter Reutemann | 1 Mar 04:24 2009
Picon
Picon

Re: IBK options

> I am troubling with calling IBK from Matlab. do you know how to set the options for using Euclidean distance?
> The option in the weka "explorer" is
>   weka.classifiers.lazy.IBk -K 1 -W 0 -A "weka.core.neighboursearch.LinearNNSearch -A
\"weka.core.EuclideanDistance -R first-last\""
> There are double quote and slash. When I put the following in Matlab, it did not work :(
>   Options={'-K' '1' '-A' 'weka.core.neighboursearch.LinearNNSearch -A weka.core.EuclideanDistance'};

I don't use Matlab, but have you tried the following?
  Options={'-K' '1' '-A' '"weka.core.neighboursearch.LinearNNSearch -A
weka.core.EuclideanDistance"'};

Theoretically, you should use the same string as the Explorer generates, i.e.:
  Options={'-K' '1' '-W' '0' '-A'
'"weka.core.neighboursearch.LinearNNSearch -A
\"weka.core.EuclideanDistance -R first-last\""'};

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

Peter Reutemann | 1 Mar 04:35 2009
Picon
Picon

Re: New to Weka need help

> I am currently using WEKA for a Computer Science Retention project.    I am
> very new to Weka and need some advice.  I successfully uploaded the training
> set and used the classify method using J48 decision tree.  I saved the model
> and apply it to the test set.  Training and Test set consists of data on CS
> students who have graduated.  The problem I am currently having is applying
> the model to future data (current CS undergraduate students).  When I apply
> the model, total number of instances classified is zero.  I am quite sure my
> post is confusing...I am confusing myself.  I can send a sample of my data
> set.

I just assume that you did something as follows on the commandline:
  java weka.classifiers.trees.J48 -t train.arff -d j48.model
  java weka.classifiers.trees.J48 -T test.arff -l j48.model

If your data test set "test.arff" does not contain any class values
(i.e., only "?") then you won't get any values in the confusion
matrix, of course, since Weka does not know the actual class value. If
you take a closer look at the output, then you will see that "Total
Number of Instances" is 0 and "Ignored Class Unknown Instances" the
number of instances in your test set.

In order to see the predictions, use the "-p" option, e.g.:
  java weka.classifiers.trees.J48 -T test.arff -l j48.model -p 0

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
(Continue reading)

Peter Reutemann | 1 Mar 04:40 2009
Picon
Picon

Re: (no subject)

> I want to now how to implement a mlp classifier by using the package
> weka.jar and not with the weka  software.
> what are the java class that form this classifiers,th method adding an
> element to this class.

If you want to implement your own classifier, have a look at FAQ "How
do I write a new classifier or filter?". Link to FAQs available from
Weka homepage.

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

Peter Reutemann | 1 Mar 04:41 2009
Picon
Picon

Re: Weka Output in Kepler

>   I am using the Weka actor in Kepler. I want to use the output of this Weka actor in Matlab, is there anyway I
can make the output compatible with matlab? (Eg. output in excel or anyway) What do I have to do?

What exact output (i.e., using what output port) do you want to use in Matlab?

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

Peter Reutemann | 1 Mar 04:57 2009
Picon

Re: Query regarding updateClassifier method with NaiveBayesUpdateable.

> I am trying to use NaiveBayesUpdateable mainly to cut down on a single huge
> set of Instances (training examples).
> Just a brief description of the problem I was facing - If I don't use
> incremental learners, I have to provide all the training data
> (by using Instances class) at the same time. This creates memory issues.
> So I thought about using NaiveBayesUpdateable so that I could split the
> training set and pass it in batches to
> overcome the memory issues. However this has resulted in decreased accuracy
> as compared to when all the data
> is passed in once (it went from 95% to 55%). I am not using any options with
> this classifier.
> Could someone throw some light on it or is there some problems in using
> NaiveBayesUpdateable ?
> regards,

How do you provide those batches? Do you perform the step-wise
training from code or commandline?

Cheers, Peter
--

-- 
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. +64 (7) 858-5174

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

(Continue reading)


Gmane