Akademik Veri Yönetim Sistemi
4th International Civil Engineering & Architecture Conference (ICEARC'25), Trabzon, Türkiye, 17 - 19 Mayıs 2025, ss.1593-1599, (Tam Metin Bildiri)
Selecting construction projects has long been a multifaceted challenge due to the complexity of variables such as cost, time, environmental impact, and strategic alignment. The critical problem identified is that existing methods in construction project selection are predominantly reliant on conventional machine learning (ML) models, which are unable to effectively capture relational dependencies among project attributes, especially for unseen or evolving data. Despite the popularity of graph convolutional networks (GCNs) in construction project management studies, these models are inherently transductive, requiring all nodes to be available during training, thereby restricting their flexibility in dynamic decision-making systems. To address these limitations, this study integrates GraphSAGE (Graph Sample and Aggregate), a state-of-the-art graph neural network (GNN) algorithm that supports inductive learning by aggregating neighborhood features, into a multi-criteria decision-making (MCDM) framework for project selection. Construction projects and investment records are represented as nodes in a graph, with edges formed based on shared regions or attributes, enabling the model to extract relational insights. Experimental evaluation on over 10,000 project investment records demonstrated that GraphSAGE achieves superior performance, with an accuracy of 92% and an F1 score of 90%, significantly outperforming GCN's accuracy of 57% and F1 score of 56%. This methodology enables enhanced decision-making by modeling project selection as a dynamic relational network, reducing decision uncertainty and improving investment strategies. The practical implications of this study include early-stage project screening, optimized resource allocation, and the development of scalable decision-support systems for construction firms, paving the way for data-driven investment in large-scale infrastructure.