Performance Assessment of Business Process Optimization Algorithms Using a Prototype Dataset Generator

Modern approaches of business administration consider processes as the fundamental element for the structure of a company that pursues high competitiveness and profitability. Hence, the implementation of effective processes has become a major concern, in an attempt to develop methodologies for their continuous improvement with respect to predefined evaluation criteria. Considering that an optimization method needs to be evaluated systematically in terms of performance and accuracy, the aim of this paper is the development of a generator that can construct synthetic test problems of diverse size and difficulty, employing well-defined complexity parameters derived from the Graph Theory and where a business process design is represented as a two-terminal-directed acyclic graph. However, a plethora of different formulations regarding the process optimization problem that have been presented in the relevant literature motivated our research to consider the target model of the generator as an ideal case, which can be derived by the aforementioned formulations after reducing some of their constraints. The experimental evaluation demonstrates how different value settings of the complexity parameters affect the difficulty of the generated instances, proving the capability of the proposed generator to create test sets of varying complexity.