CSCE 633 
Project 3 - Nearest Neighbor
due: Tues, Apr 10, 2012

Implement a nearest-neighbor classifier and compare its
performance to your decision tree and neural network on 
at least 4 datasets from the UCI repository.

You will have to make decisions about the best approach, such as
whether to use k-NN with k=3, k=5... or to use distance-weighting,
etc.  I recommend implementing and testing several approaches to see
which is best.

You will have to decide how to combine symbolic and numeric
attributes in a distance formula. (There are several methods
we discussed in class.  Also, don't forget to normalize the
input data values.)

You might want to try condensing or editting (though this is not
required).

An additional component of this project to is to implement and test
some form of feature weighting or feature selection (your choice).
Report which features were selected or got the highest weights.

For comparing to your other classifiers, choose the best version of
each.  Conduct a statistical comparison of the best version of each of
the 3 algorithms, and focus on trying determine which is best (it may
depend on the database).  Explain your findings.

What to turn in: a written report, plus a print-out of your code.