A GP Process Mining Approach from a Structural Perspective

Abstract

Process mining is the automated acquisition of process models from event workflow logs. And the model’s structural complexity directly impacts readability and quality of the model. Although many mining techniques have been developed, most of them ignore mining from a structural perspective. Thus in this paper, we have proposed an improved genetic programming approach with a partial fitness, which is extended from the structuredness complexity metric so as to mine process models, which are not structurally complex. Additionally, the innovative process mining approach using complexity metric and tree based individual representation overcomes the shortcomings in previous genetic process mining approach (i.e., the previous GA approach underperforms when dealing with process models with short parallel and OR structure, etc). Finally, to evaluate our approach, experiments have also been conducted.