An optimization-based process mining approach for explainable classification of timed event logs

This paper addresses the problem of supervised classification of time event logs of two classes: positive and negative population. The key idea of this paper to explain classification is to determine some process model that fits well the positive event logs and poorly the negative ones. More specifically, we introduce formal definitions of event logs, process models and a replayability score that measures the fitness of a process model for a given event log. We then set the event log classification as an optimization problem for the determination of a process model that maximizes its replayability for the positive population and minimizes its replayability for the negative one. A tabu search algorithm is then proposed to solve this problem. The proposed algorithm is compared with three state-of-the-art classification algorithms on test cases of various complexity. It is shown to provide superior performances and a graphic representation of the process model of the positive event logs.