Optimizing the Solution Quality of Metaheuristics Through Process Mining Based on Selected Problems from Operations Research

Methods from Operations Research (OR) are employed to address a diverse set of Business Process Management (BPM) problems such as determining optimum resource allocation for process tasks. However, it has not been comprehensively investigated how BPM methods can be used for solving OR problems, although process mining, for example, provides powerful analytical instruments. Hence, in this work, we show how process discovery, a subclass of process mining, can generate problem knowledge to optimize the solutions of metaheuristics to solve a novel OR problem, i.e., the combined cobot assignment and job shop scheduling problem. This problem is relevant as cobots can cooperate with humans without the need for a safe zone and currently significantly impact transitions in production environments. In detail, we propose two process discovery based neighborhood operators, namely process discovery change and process discovery dictionary change, and implement and evaluate them in comparison with random and greedy operations based on a real-world data set. The approach is also applied to another OR problem for generalizability reasons. The combined OR and process discovery approach shows promising results, especially for larger problem instances.