Hybrid GA-manfis model for organizational competencies and performance in construction

The majority of competency and performance modeling methods available in the literature are deterministic conceptual, statistical, and/or regression models that cannot capture the subjective uncertainty, complex, and nonlinear relationships inherent in construction, which makes accurate prediction difficult. Past studies utilized neuro-fuzzy system (NFS) models, such as adaptive neuro-fuzzy inference system (ANFIS), that combine the learning power of artificial neural networks and functionality of fuzzy systems to develop accurate predictive models. ANFIS is robust, fast, and effective in solving complex problems for a range of real-world construction engineering and management (CEM) applications. NFS models such as ANFIS have some limitations in handling multiple outputs common in construction industry problems, such as being prone to early convergence due to local minima entrapment. To address these limitations, this paper proposes a hybrid NFS combining the evolutionary optimization technique of a genetic algorithm (GA) with a multi-output adaptive neuro-fuzzy inference system (MANFIS) that can handle multi-input multi-output (MIMO) problems for CEM applications. The proposed modeling approach is demonstrated using a case study that showed good results in predicting multiple organizational performance metrics using organizational competencies. The contributions of this paper are threefold: It (1) proposes a novel methodology of integrating different computing techniques for developing a GA-based multi-output adaptive neuro-fuzzy inference system (GA-MANFIS) model that can handle complex and nonlinear MIMO problems inherent in construction processes and practices; (2) relates organizational competencies to performance and predicts multiple organizational performance metrics; and (3) provides a GA-based feature selection approach that reduces data dimensionality, enabling identification of organizational competencies that significantly influence organizational performance. By uniquely integrating these techniques, this model enables construction organizations to evaluate their competencies and predict multiple organizational performance metrics simultaneously, and researchers can adapt it for a variety of construction contexts.