CS 497 Assignment -- Models for Prediction

Gordon V. Cormack

Use a data compression model to do perform one of the following tasks:

classification (e.g. spam filtering, authorship detection, classification by language)
game playing (e.g. rock-paper-scissors, liar's dice, ...)
some other statistical prediction task

Design and execute an experiment that measures the effectiveness of your approach. In doing so you will have to acquire some data (e.g. some email from your inbox or the result of playing the game) and measure the result. Make sure that the data used to measure the result plays no role in the design or training of our method.

Your approach need not be grandiose. It is perfectly acceptable, for example, to collect a few dozen messages, split them into training and test sets, and use the method described in class. You may use any compression model you like, including DMC, PPM, or any popular compression software.

You should write one or two pages describing your method, the experiment you designed including the source of the data, and measured results. Your write-up should include one paragraph of observations and discussion.