Posted by: zyxo | December 30, 2007

Swarm Intelligence and Data mining


Anyone familiar with advanced data mining techniques knows that the best models are obtained with techiques relying on splitting up the rules into many tiny pieces and afterwards look for the ‘emerging’ general rule by sort of averaging up all the little bits.

Examples of this superority are the contests won by Salford products, and currently the best results in the netflix contest. Another example can be found
here
.

Among these ’swarm-like’ data mining technique : artificial neural nets, bagging, boosting, random forests, treenets, ant clustering (which is still experimental).

The pattern is still the same and somewhere the definition of swarm intelligence : the intelligence of the swarm is allways much higher than the intelligence of the items in the swarm.

The drawbacks of these complex techniques/models are firstly that they very much like a black box : it works, but you do not kwow why or how and secondly they consume a lot of computer resources.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: