EFFICIENCY ANALYSIS OF FUZZY LOOP DETECTION METHODS

Abstract

Automated Process Discovery is aimed at determining the most fitting business process schema model from the event logs. One of the most difficult schema components is the loop structure and current process mining methods lack efficient loop mining. Our paper presents two novel approaches to fuzzy loop mining, where in fuzzy loops, the repeating kernel may have different variations. On of the proposed methods uses item-wise direct comparisons, while the other one is based on CNN neural network architecture. The performed evaluation tests show which are the benefits or drawbacks of the different loop mining approaches providing a guide for the implementation of efficient schema mining frameworks.