3
votes
$\begingroup$

I am doing a project with Weka with clustering algorithms (simple k-means, expectation-maximization and OPTICS). I would love to be able to call those algorithms over the internet.

So does Weka has a online API? But I suspect there is no. In that case, what are the alternatives?

Edit: What I am trying to achieve is a web interface to launch and control clustering algorithms. But I think I will end up running Weka directly on the server, either by invoking command line, or using my own java program that uses Weka.

Another advantage would have been not to be limited by the server hardware but this is acceptable in my case.

I have seen that Apache Mahout has a large list of algorithms. Apache Mahout is a library of machine learning algorythms, but also without a callable online API.

Google has the kind of "machine learning as a service" platform I want which is the Google Prediction API, but it has more real world algorythms like spam detection or email classification, and not the clustering algorythms I want.

$\endgroup$
3
  • $\begingroup$ WEKA does not have online APIs. Any particular reason why you want to do it over the internet? It will be useful if you describe what you are doing and why. $\endgroup$
    – user721975
    Commented Dec 7, 2011 at 23:09
  • $\begingroup$ Could you clarify what exactly you are trying to do? It sounds like you want Weka to run on a server and for your client to be able to call clustering algorithms on that server. Is that right? $\endgroup$
    – Michael McGowan
    Commented Dec 7, 2011 at 23:14
  • $\begingroup$ Edited my question $\endgroup$
    – Benjamin Crouzier
    Commented Dec 8, 2011 at 0:32

2 Answers 2

Reset to default
4
votes
$\begingroup$

RapidAnalytics, a cousin of the RapidMiner project, does exactly that.

You model your process in RapidMiner, using Weka or native RapidMiner operators, and then publish it to the RapidAnalytics server where it is available as a web service, returning XML, HTML or JSON to you.

It's impressive open source technology with a great community.

http://rapid-i.com/content/view/182/192/lang,en/

(There are enterprise versions available, but what you need can be done with the open source version.)

$\endgroup$
2
  • $\begingroup$ (+1) impressive. Questions: 1.Rapid Analytics is completely hosted by rapid-i ? 2. If the 1. is true, do you know the resource limitations (cpu time, ram etc.) an open-source user has to expect ? $\endgroup$
    – steffen
    Commented Dec 8, 2011 at 15:39
  • $\begingroup$ RapidAnalytics is also an open source package for installing on your server. $\endgroup$
    – Balázs Bárány
    Commented Dec 9, 2011 at 6:21
0
votes
$\begingroup$

OPTICS in Weka is pretty much an external application. It's student code an unmaintained. Weka is good for classification, anything else is of very limited use.

For clustering, I'd look at different projects, for example ELKI, which has a much better and much faster OPTICS implementation. In general, for clustering and outlier detection it is miles ahead of Weka - kind of the "Weka for clustering and outlier detection".

$\endgroup$
0

Not the answer you're looking for? Browse other questions tagged or ask your own question.