DM: Introductions

[Ronny Kohavi]

Hello everyone,

My name is Ronny Kohavi.  I manage the analytical data mining group
that writes the core mining algorithms for MineSet, Silicon Graphics'
data mining product (http://www.sgi.com/Products/software/MineSet).

The algorithms in MineSet are based on the MLC++ project, the machine
learning library in C++, that I headed at Stanford University and
which is distributed in source code through Silicon Graphics
(http://www.sgi.com/Technology/mlc).

My main interests include fielding the machine learning technology,
which we are doing well now through MineSet.  The MineSet group
concentrates on generating models that are interpretable and can be
visualized (e.g., our decision tree visualizer provides fly-throughs,
drilling, etc).  We believe that for the knowledge discovery process
to happen quickly, users need to look under the hood to see what the
models test and visualization is a key component of providing users
with that information.

We are currently looking at new ways to visualize models, and if you
have any ideas we're very interested.  For example, how would you
visualize a nearest-neighbor model?  How would you visualize 10,000
rules to a user?

--

   Ronny Kohavi (ronnyk@sgi.com, http://robotics.stanford.edu/~ronnyk)