Feature selection and democratic consensus metrics in large-scale group decision-making: A methodological integration

In the era of Big Data, managing complex and vast datasets while extracting informed decisions has become a critical challenge, especially in the context of artificial intelligence (AI) applications in engineering. Large-scale group decision-making (LSGDM) has thus emerged as a key area of research in decision science. Dimensionality reduction and consensus models are two essential components of LSGDM. Although cluster analysis is widely used for dimensionality reduction, it often falls short in fully reducing the complexity for decision-makers (DMs) in engineering settings. To address this, we propose a novel feature selection method based on ranked cluster analysis, which filters out redundant decision information, significantly reducing the dimensionality for DMs and the complexity of consensus models. We introduce a democratic consensus metric rule that allows DMs to independently select consensus metrics that best suit their interests, enhancing overall consensus levels and reducing consensus-building costs. Additionally, we design a feedback mechanism to ensure efficient convergence to a final consensus. The effectiveness of our method is validated through real-world cases from the UCI database and a series of comprehensive experiments, including comparative, sensitivity, simulation, and ablation analyses. These tests highlight the robustness and practical applicability of our approach in complex decision-making scenarios. Our research contributes a novel and efficient solution to LSGDM, particularly relevant in AI-driven engineering, environmental management, and policy-making.