FR3 - Fuzzy Round Robin Ripper
The FR3 algorithm is a pairwise classifier which is based upon the Ripper algorithm (by Cohen) in a pairwise learning environment, also called R3 (by Fürnkranz). The FR3 system extends the R3 approach by fuzzy techniques and structured preferences. Besides the improvement in classification accuracy in comparison to R3, the FR3 is able to express conflicts and ignorances in a classification decision in an even more faithful way.
On this page you will find the implementation of the FR3 algorithm, which runs in the WEKA environment. In the archive file (see below) you will find the source code, the documentation and a directly runnable version of FR3 (combined with WEKA 3.5.8.). To download datasets for testing, please refer to the WEKA-homepage.
Note: This is the latest version of FR3, made public on August 7th, 2008. It supports WEKA 3.5.8, was slightly improved and offers an extension to visualize a fuzzy preference structure.