TY - JOUR TI - A strategy for evaluating the quality of trace alignment tools based on a Markov model DO - https://doi.org/doi:10.7282/T3M32T7J PY - 2014 AB - Trace alignment of event logs is used to understand and improve business processes. A key missing component of current approaches for performing trace alignment is a methodology to measure the quality of alignment. We propose a novel approach for generating random event logs that can be used for testing and evaluating trace alignment tools. We first extracted a statistical model from 437 real-world traces from Children’s National Medical Center in Washington DC (CNMC). We then created a guide tree and prune it to a minimum spanning tree based on user defined trace number in output event log. The final step is to fill this tree. Each node in this tree contains a trace. Each leaf node represents a trace in output event log. The root node is filled by a user-defined sequence and each child node is mutate from parent node based on the statistical model. To validate our approach, we used a concept of replay fitness score. Replay fitness score is used to quantify the extent to which a model can reproduce the traces recorded in an event log. It’s between 0 and 1. The value 1 means that the model can perfectly replay the event log and 0 means that the model cannot reply the log. Comparing with process model (Petri-Net) extracted from 437 real-world traces, the output event log of our system can constantly get a score of 0.8. II Therefore, our results are relevant not for only validation of trace alignment tools but also for other process mining tools. KW - Electrical and Computer Engineering KW - Markov processes LA - eng ER -