Xiaonan Li

Existing large-scale group decision-making (LSGDM) models often rely on forced rankings when processing hybrid online reviews, which may overlook inherent uncertainties and lead to deterministic decisions that ignore misclassification risks. This paper proposes a risk-aware LSGDM framework to address this issue. First, textual and numerical reviews are unified into a basic uncertainty information (BUI) structure. A BUI-driven consensus process is then developed to manage group disagreements while preserving original individual opinions. To facilitate risk-aware decision-making, a rigorous metric space is required for data-driven threshold derivation. Thus, we establish an uncertainty-measure-preserving mapping between the BUI algebraic space and the intuitionistic fuzzy number (IFN) geometric space. This mapping enables the integration of granular-ball computing with three-way decision (TWD) theory, allowing for the adaptive generation of decision thresholds. Experiments on real-world datasets illustrate the feasibility and reliability of the proposed framework.