PDA

View Full Version : Weka - where to start?



codek
02-18-2008, 12:09 PM
Hi,

What exactly is weka? Or; more relevant to my requirements, can it analyse a stream of data and find patterns and/or trends? If so how do i do that, which bit of the tool etc? Experimenter? Explorer? Eh?

If so how do you start? I can find very little documentation, and there's no quick start guide. Is it the sort of tool you can just throw some data into and see what comes out?

I want to be able to get my head around what it is and what it can do... I can see it's updated fairly often, but the lack of posts in here imply to me that it's rarely used?

Thanks,
Dan

Mark
02-18-2008, 04:46 PM
Hi Dan,

Weka is a workbench of machine learning/data mining algorithms + GUIs. It can be used to find predictive patterns in data (i.e. mappings between input fields and a target field of interest) and for clustering (i.e. segmenting data into "natural" groups).

Machine learning/data mining is an advanced area of analytics. It is usually not the case that a point-and-click/throw some data at the problem type of approach will yield good results. If you are new to data mining and Weka, I'd suggest taking a look at the book that was written by a professor and a senior lecturer from University of Waikato to accompany the Weka software:

Ian H. Witten and Eibe Frank. 2005. Data Mining: Practical Machine Learning Tools and Techniques. Morgan Kaufmann.

http://www.amazon.com/Data-Mining-Practical-Techniques-Management/dp/0120884070/ref=pd_bbs_sr_1?ie=UTF8&s=books&qid=1203367217&sr=8-1

There is documentation on Weka (including its various user interfaces) available both online and included (in pdf form) in the Weka distribution when you download it. See:

http://weka.pentaho.org
http://wiki.pentaho.org/display/DATAMINING/Home
http://www.cs.waikato.ac.nz/ml/weka

HTH.

Cheers,
Mark.