Evolutionary Business Process Optimization using a Multiple-Criteria Decision Analysis method

In this article, a research was conducted on the importance of the initialization of population on the solutions resulting from an evolutionary multi-objective optimization (EMOO). This means how the technique applied in the population's initialization can affect -supposed- optimized solutions after a certain number of iterations in a multi-objective optimization using evolutionary algorithms (EAs). Knowing that Pareto front is the ideal to reach, the presented work propose to utilize a Multiple-Criteria Decision Analysis method (MCDA) to sort or classify the initial population. To achieve this goal, an EMOO Framework is implemented which uses (i) conflicting optimization criteria, (ii) NSGAII as an EA and (iii) MR-sort method is used as a MCDA for classification. The experiments results clearly indicate that the proposed Framework is capable of producing an acceptable number of optimized design alternatives regarding the problem complexity and in a reasonable period.