Check C4.5's pruning operations. Specify different rates and see how testing accuracy changes on iris and usps (for iris, you can randomly split the date to 2/3 training and 1/3 testing). You can try extreme rates to see and explain what happens.
Trace the code and explain how it does the pruning operation. You can use weka, r-weka, or any implementation of C4.5.
Write a short report (<= 2 pages in English).