CS 205 Final Project:
You may work alone or in groups of two. This is due on Friday at 5:00pm of finals week.
1) Implement the K-nearest neighbor algorithm (my version of the 1-nearest neighbor algorithm in
matlab takes 10 lines of code, this should not be too tricky). Download some datasets to test on [a]
[b] (try to focus only on larger datasets).
2) Using cross validation, investigate the effect of changing the “K” in K-nearest neighbor algorithm.
3) It is likely that for some datasets, some attributes are irrelevant. Using some kind of search, try to
find a subset of instances that will produce improved accuracy.
Possible search algorithms for attribute selection include: forward selection, backward selection, genetic
algorithms, random search, heuristic search etc
Write a report for the above, in conference paper format (see [c] as an example). I expect it will take 3 to 5
pages.
[a] http://kdd.ics.uci.edu/
[b] http://archive.ics.uci.edu/ml/
[c] http://www.cs.ucr.edu/~eamonn/Anytime_Classification.pdf
