Differences

This shows you the differences between two versions of the page.

Link to this comparison view

blogs:pub2016:the_drfurby_classifier_submission_to_the_process_discovery_contest_bpm_2016 [2016/09/30 11:13] (current)
hverbeek
Line 1: Line 1:
 +====== The DrFurby Classifier submission to the Process Discovery Contest @ BPM 2016 ======
 +
 +H. M. W. Verbeek and F. Mannhardt, [[http://​bpmcenter.org/​wp-content/​uploads/​reports/​2016/​BPM-16-08.pdf|The DrFurby Classifier submission to the Process Discovery Contest @ BPM 2016]], BPMCenter.org,​ BPM Center Report BPM-16-08, 2016.
 +
 +===== Abstract =====
 +
 +To check the current state-of-the-art in process discovery, a
 +contest has been set up. In this contest, 10 undisclosed process models
 +were created by the organizers. For every created process model, the
 +organizers created two logs: a training log containing 1000 traces and a
 +test log containing 20 traces. Both logs were disclosed to the participants.
 +The challenge for the participants was to classify every trace from every
 +test log whether or not it  ts the corresponding original process model,
 +that is, whether or not it can be replayed perfectly by that process model.
 +As the original process model was not disclosed, the participants had to
 +first discover an as-good-as-possible process model from the training log,
 +and second use this discovered process model to classify every trace from
 +the corresponding test log. The main assumption here is that a better
 +discovery technique discovers a better (that is, more matching to the
 +original) process model, which then results in a better classification. In
 +the end, the one that classifies the most traces correctly, wins the contest.
 +This paper presents a submission to this contest that, in the end, classifies
 +193 out of 200 traces correctly.
 +