Effect of seemingly unrelated regression-based modeling approach on solution quality for correlated multiple response optimization problems

Multiple response optimization remains a critical and important research area in quality engineering and management. Various methodologies have been proposed to resolve a correlated multiple responses optimization problem. However, very few address the importance of empirical response surface modeling and its influence on the optimal solution quality. In this paper, two different approaches of empirical modeling, using multiple regression, viz. ordinary least square (OLS), and seemingly unrelated regression (SUR) are selected for study. To compare the approaches, two different metaheuristic optimization strategies are used, viz. ant colony optimization in real space (ACO_R) and Honey Bee Optimization algorithm (HBO) for a given case situation. Two different cases illustrate that SUR-based response surface models provide significantly better solution than OLS approach for correlated multiple response problems.