Optimizing Resource-Driven Process Configuration Through Genetic Algorithms

In order to optimize the efficiency of operations in organizations, the control flow of business processes and the resources allocated to process tasks have to be considered in an intertwined way. In real-world process scenarios, resources might even manipulate the control flow, e.g., if the allocation of a certain resource to one task renders the execution of another task superfluous. Hence, we advocate to equip resources with change patterns, resulting in process configuration at instance level. This raises the challenge of determining executable process configurations with valid and, at the same time, optimal resource allocations w.r.t. some optimization goal. To this end, we introduce and utilize the concept of the Resource-Augmented Process Structure Tree (RA-PST) with insert, replace, and delete patterns for resources. The RA-PST combines the variability of configurable process models with optimization-focused resource allocation modeling. It is shown how the validity of the resource allocation and the soundness of the resulting process instance can be checked based on the constructed RA-PST. For the combinatorial optimization problem of resource allocation, we adopt a genetic algorithm and test it on five different sets of resources. The results showcase the effectiveness of focusing on resource optimization during business process modeling and demonstrate how an optimal configuration can be achieved, i.e., the genetic algorithm finds (near-) optimal solutions, especially when heuristics are not able to handle the additional complexity.