Optimal process mining of timed event logs

Abstract The problem of determining the optimal process model of an event log of traces of events with temporal information is presented. A formal description of the event log and relevant complexity measures are detailed. Then the process model and its replayability score that measures model fitness with respect to the event log are defined. Two process models are formulated, taking into account temporal information. The first, called grid process model, is reminiscent of Petri net unfolding and is a graph with multiple layers of labeled nodes and arcs connecting lower to upper layer nodes. Our second model is an extension of the first. Denoted the time grid process model, it associates a time interval to each arc. Subsequently, a Tabu search algorithm is constructed to determine the optimal process model that maximizes the replayability score subject to the constraints of the maximal number of nodes and arcs. Numerical experiments are conducted to assess the performance of the proposed Tabu search algorithm. Lastly, a healthcare case study was conducted to demonstrate the applicability of our approach for clinical pathway modeling. Special attention was paid on readability, so that final users could interpret the process mining results.