Group Presentations:
I present four different topics here for group presentations. I provide
**some** references and ask that you find others. You should plan on
using a class period for a class presentation. Your group should also
submit a small (5 to 10 page) paper.
What you need to do soon : form a group of 1 to 3 people and tell me
who is in your group!
1) Greenhouse gases and global
warming
A big topic here!! Let's avoid the politics and stick to
the facts.
Many suggest that carbon dioxide emissions are a major cause of global
warming. A well known paper supporting this view is from Mann [MNB98]
where the famous "hockey stick" proposal is mentioned. Several
scientists cite error with the methods reported in this paper, claiming
that Mann's use of Principal Component Analysis would cause a
"spike" in any data set (and not only data with CO2 data). Needless to
say, the critics have themselves been criticized. OK. Let's find some
references here:
Names to do google searches on: Mann, McIntyre
and McKitrick
Canadian critics of the Mann report can be found here
Some more critics here
Web site of climatologists supporting the idea of greenhouse gas
warming can be found here
. Do a search on "hockey stick" at this site for some interesting
reading.
There is some pretty techincal reading here. Does either side do
anything you find questionable ?
2) Data Mining and Wall Street
Investing
Again, another big topic. There are many
papers (and ideas!!) concerning data mining applications to successful
investing, but here's a cute one I think is worth looking at. Several
authorities (and I provide two here) suggest that the two weeks of the
new moon, that is, the seven days leading to the new moon and the
subsequent seven days following are better times to invest!! How did
the researchers (some from University of Michigan) come up with this ?
Lunar
Cycle Effects in Stock Returns
Are
Investors Moonstruck
3) Logistic Regression and e-mail
SPAM
We
have already seen application of Naive Bayesian inference to
spam detection. Recent work suggests that logistic regression is
better; its easier to train and faster. So what is logistic regression
and how can we use it for spam detection. Do a google search on
logistic regression. A reasonable place to start is here.
Joshua
Goodman of Microsoft research has many papers related to logistic
regression and spam. Another place to look is at this conference web site
, again, scan the page for logistic regression.
4) Social Network Analysis and
Terrorists
The President wants to tap your phone. Not really...but he
wouldn't mind keeping track of what numbers you call and what numbers
call you. It is felt that data mining of social networks can provide
useful information in finding terrorists. Here are some interesting
articles:
Can
Network Theory Thwart Terrorists
Social Network Analysis
of the 9/11 Terrorist Network
Nice reference Page for Social
Network Analysis