Are Decision Trees Always Greener on the Open (Source) Side of the Fence?

- This short paper compares the performance of three popular decision tree algorithms: C4.5, C5.0, and WEKA's J48. These decision tree algorithms are all related in that C5.0 is an updated commercial version of C4.5 and J48 is an implementation of the C4.5 algorithm under the WEKA data mining platform. The purpose of this paper is to verify the explicit or implied performance claims for these algorithms--namely that C5.0 is superior to C4.5 and that J48 mimics the performance of C4.5. Our results are quite surprising and contradict these claims. This is significant because existing work that is based on these claims (e.g., J48 being equivalent to C4.5) may be misleading.

• Type: Journal
• Year: 2009
• Where: DMIN
• Authors: Samuel Moore, Daniel D'Addario, James Kurinskas, Gary M. Weiss
• Keywords: Algorithms, C4.5, Data Mining, Decision Tree, DMIN 2009