https://noisebridge.net/index.php?title=Machine_Learning_Meetup_Notes:_2010-06-09&feed=atom&action=historyMachine Learning Meetup Notes: 2010-06-09 - Revision history2014-04-20T07:24:37ZRevision history for this page on the wikiMediaWiki 1.19.1https://noisebridge.net/index.php?title=Machine_Learning_Meetup_Notes:_2010-06-09&diff=11702&oldid=prevSpammerHellDontDelete: Created page with 'Andy's final features: *problem hierarchy - 200 nominal *iq - percentage of getting it right on the first try, numeric *iq strength - number of those, numeric *chance - …'2010-06-10T04:00:20Z<p>Created page with 'Andy's final features: *problem hierarchy - 200 nominal *iq - percentage of getting it right on the first try, numeric *iq strength - number of those, numeric *chance - …'</p>
<p><b>New page</b></p><div>Andy's final features:<br />
<br />
*problem hierarchy - 200 nominal<br />
*iq - percentage of getting it right on the first try, numeric<br />
*iq strength - number of those, numeric<br />
*chance - numeric<br />
*chance strength - how hard the step is - percentage of students who got the step right, numeric<br />
*num subskills - numeric<br />
*num tracedskills - numeric<br />
*predict -> cfa - percentage<br />
<br />
all the numerics were discretized and run through naive bayes, that gave a 0.38 root mean squre error (MSE)<br />
<br />
he stuck in a .87 for all of them and got a 0.33 MSE<br />
<br />
final method: cfa = iq + 2(chance)/3 gave 0.31 MSE<br />
<br />
<br />
Mike: presents clustering with graphs<br />
<br />
*JUNG - java tool for graph viz and contruction<br />
*eliminate edges that are the heaviest<br />
*minimum spanning tree<br />
*nb-sourceforge project<br />
*weight represents euclidean distance of two points<br />
*similar to pricipal component analysis</div>SpammerHellDontDelete