tag:blogger.com,1999:blog-29117034.post1493846093426071201..comments2024-01-08T12:39:20.176+01:00Comments on Data Mining Research - dataminingblog.com: What have you mined?Sandro Saittahttp://www.blogger.com/profile/17682082649770875583noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-29117034.post-29467145947429266782008-01-08T17:53:00.000+01:002008-01-08T17:53:00.000+01:00Thanks for your relevant comments Will. Regarding ...Thanks for your relevant comments Will. Regarding images, Google is certainly using data mining algorithms for its SafeSearch filtering. Up to now, my favorite is definitely the "popcorn" one!Sandro Saittahttps://www.blogger.com/profile/13661800307652043117noreply@blogger.comtag:blogger.com,1999:blog-29117034.post-5569859747086785552008-01-08T16:06:00.000+01:002008-01-08T16:06:00.000+01:00On a more puerile note, there have been a number o...On a more puerile note, there have been a number of research projects which attempted to, in turn: 1. discern human skin among pixels in images, 2. identify unclothed human bodies in images (a "pornography detector"), and 3. assess human facial beauty. Some examples are:<BR/><BR/><A HREF="http://cs-people.bu.edu/ruel/cs585/HW1/jwortman/hw1/hw1.html" REL="nofollow">Skin Detection, by Jennifer Wortman</A><BR/><BR/><A HREF="http://citeseer.ist.psu.edu/532015.html" REL="nofollow">Skin Patch Detection in Real-World Images, by Hannes Kruppa, Martin A. Bauer and Bernt Schiele</A><BR/><BR/><A HREF="http://www.stanford.edu/class/cs229/proj2006/HefnerLindsay-AreYouHotOrNot.pdf" REL="nofollow">Are You HOT or NOT?, by Jim Hefner and Roddy Lindsay</A><BR/><BR/>Whatever their other attractions, these problems are technically interesting. Human skin, for instance, can be difficult to discern from things like wood grain.Will Dwinnellhttps://www.blogger.com/profile/03379859054257561952noreply@blogger.comtag:blogger.com,1999:blog-29117034.post-23957561569929664282008-01-05T18:51:00.000+01:002008-01-05T18:51:00.000+01:00Professionally, most of my predictive modeling wor...Professionally, most of my predictive modeling work has been pretty conventional, with things like:<BR/><BR/>-customer attrition (telecommunications and mutual fund customers)<BR/>-credit risk<BR/>-business forecasting<BR/>-industrial part quality prediction<BR/>-medical diagnosis (cancer)<BR/><BR/>Recreationally, I'll predict anything for fun. A good example is my <A HREF="http://matlabdatamining.blogspot.com/2007/01/pixel-classificiation-project.html" REL="nofollow">Pixel Classification Project</A>, in which individual pixels in images are classified as "foliage" or "not foliage".<BR/><BR/>I have met people or read about the following predictive modeling applications:<BR/><BR/>-classifying popcorn kernels as to whether or not they will pop<BR/>-estimating lumber value of plots of land from aerial photographs<BR/>-assessing drug toxicity from images of deformed webs of spiders given said drugs<BR/><BR/>Of course, there are myriad examples, both on-line and in more formal settings, of people attempting to predict things like the stock market, horse races, etc., though most of the ones I've examined closely were not sufficiently rigorous to be very inspiring.Will Dwinnellhttps://www.blogger.com/profile/03379859054257561952noreply@blogger.com