[kepler-users] How to integrate Weka into Kepler?

[Dan Higgins]

Hello again Josep,
    I took a quick look at Weka and have a few more comments.

    First of all, there is apparently a command-line interface version 
of Weka. This means that it can be run in Kepler using the command line 
actor. Basically, anything you can do on the comand line should be 
doable using the CommandLine actor.
    It also looks like there is an existing GUI interface, so you could 
just launch that interface inside Kepler again using the Command line actor.

Dan Higgins

Josep Maria Campanera Alsina wrote:
> Dear Community,
> I'd like to integrate Weka 3.5.7 into the Kepler environment. Weka is
> the most well-known data mining library. It is written in Java
> http://www.cs.waikato.ac.nz/ml/weka/.
> I consider this software extremely useful and it would add numerous
> capabilities to Kepler, like R did! Weka includes: data sources,
> Datasinks, filters, classifiers, clusterers, associations, evaluation
> and visualization tools! to sum up, It is another piece of jewellery
> like kepler!
>
> Well, there several ways to do that:
> (1) Using the RWeka project.
> http://cran.r-project.org/web/packages/RWeka/index.html
> Then the Weka algorithms can be called like R commands in the R
> console actor of Kepler. However it doesn't seem the more powerful way
> since Weka algorithms are also designed to run under a workflow
> environment (KnowledgeFlow environment).
>
> (2) Fully integrated: each Weka algorithm is converted to a Kepler
> actor. Here comes my advice request. Which is the more efficient way
> to undertake that process? Weka is open source software issued under
> the GNU General Public License. So, java and class files for each
> algorithm are completely available. Since I'm not a computer scientist
> the only idea that comes to my mind is to embed/add the Weka java code
> into the standard anatomy of a java Kepler actor code (ports,
> parameters, action methods ...). Definitely this is a very tedious and
> never ending task (it has to be done individually for each algorithm).
> An ideas how to tackle that? Anyone can envisage a faster and more
> systematic alternative?
>
> Best wishes,
>
> Josep Maria,
>
>
> ---------------------------------------------
> Josep Maria Campanera Alsina
> Juan de la Cierva Researcher
> Departament de Fisicoquímica
> Facultat de Farmàcia
> Avgda Joan XXIII, s/n
> 08028 Barcelona · Catalonia · Spain
> Tel: +34 93 4035988
> Fax: +34 93 4035987
> campanera at ub.edu
> --------------------------------------------
> _______________________________________________
> Kepler-users mailing list
> Kepler-users at ecoinformatics.org
> http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/kepler-users
>
>