CSCE 633 
Project 4 - Naive Bayes
due: Thurs, Apr 26, 2012

In this project, your objective is to implement the Naive Bayes (NB)
algorithm (in any language you like), test it on at least 4 datasets
from the UCI Machine Learning Repository, and compare it to your
other classifiers (decision tree, neural net, and nearest neighbor).

In NB, you make the assumption that attributes are independent (though
this is not always true in practice, and may lead to underperformance
on certain data sets).  You will have to decide how to handle both 
continuous and discrete attributes at the same time.  For continuous
attributes, you might want to fit them with an Normal distribution.
For discrete, you will want to tabulate the conditional probabilities
of classes with individual attribute values.  Remember that it might
help to use m-estimates to avoid 0's in cases where data was not
observed.

On which databases does NB work well?  On which does it perform
poorly, any why?  Be sure to use appropriate statistics to backup
you claims and observations.

Also, in this report, include a summary section that compares all of 
your algorithms together and draws some conclusions about which is
best (if any) and why (give some interpretation).

(Just a note: if you wanted to capture some *dependence* between
variables, you could construct a Bayesian network with directed edges
between variables that influence each other.  Then you could use a
learning algorithm to learn the parameters in the network (conditional
probabilies on each edge). However, that is beyond the scope of this
project.)