A Graph and Trace Clustering-based Approach for Abstracting Mined Business Process Models

摘要

Process model discovery is a significant research topic in the business process mining area. However, existing workflow discovery techniques run into a stone wall while dealing with event logs generated from highly flexible environments because the raw models mined from such logs often suffer from the problem of inaccuracy and high complexity. In this paper, we propose a new process model abstraction technique for solving this problem. The proposed technique is able to optimise the quality of the potential high level model (abstraction model) so that a high-quality abstraction model can be acquired and also considers the quality of the sub-models generated where each sub-model is employed to show the details of its relevant high level activity in the high level model.