CSCE 633
Machine Learning (Spring 2010)
due: Tues, Apr 20

Project 4
---------

Goal: Implement and test a Naive Bayes learner, and compare
its performance to your decision-tree, neural network, and nearest
neighbor algorithm on 5 different datasets.

This is a relatively straightforward project.  Mainly, you just have
to compute statistical summaries of your training data (prior and
conditional probabilities), and then apply them to test examples using
the formula in the book (based on Bayes Rule + Independence
Assumption).  For continuous attributes, model them as a
normally-distributed.  (Optionally, you might want to compare this to
a simpler approach of discretizing continuous attributes into k bins
and treating them as nominal; you could compare the performance to
using Gaussian distributions.)  For nominal attibutes, you might want
to try one of several approaches we talked about in class, or perhaps
compare two approaches to see which is best.