Akademik Veri Yönetim Sistemi
ENGINEERING MANAGEMENT JOURNAL, cilt.33, sa.1, ss.30-44, 2021 (SCI-Expanded)
Although conceptualized in various ways within the project management literature, complexity is usually identified as one of the drivers of uncertainty, which may result in deviations from the expected project performance in the construction industry. Effective strategies to manage risk and complexity can only be formulated by understanding the complexity-uncertainty-performance triad in construction projects and modeling the non-linear interactions between these factors. In this study, a meta-modeling approach was utilized for understanding complex interactions within the triad. Using a database of projects from the seed scenarios developed by experts, a Bayesian Belief Network (BBN) meta-model was constructed for capturing complex interactions across the triad. Shapley value from Game Theory was adapted to assess the output of the BBN meta-model for establishing the relative contribution of complexity and uncertainty factors to project performance while considering the efficacy of potential strategies. The output of the Shapley value analysis was used to develop an Artificial Neural Network (ANN) model for predicting project performance. The application of the proposed methodology was demonstrated through a case study involving experts from an international construction company. The results clearly manifest the merits of integrating these approaches in a unique order in that BBNs are useful for understanding the holistic account of complex interdependencies involved in construction projects, utilization of Shapley value helps in establishing the relative importance of factors across a BBN meta-model and allocating limited resources to manage project complexity and uncertainty, and ANNs exhibit an improved accuracy in predicting the project performance given the relative importance established across complexity and uncertainty factors. The findings from the case study demonstrate that the proposed methodology can be used to understand how performance may change with respect to different combinations of implemented strategies, complexity and uncertainty factors.