Abstract: As an essential component of smart city governance, disaster emergency management requires rapid and reliable decision support, yet it is often constrained by substantial practical challenges. To address these issues, a similarity-trust driven heterogeneous large-scale group decision making framework for emergency response planning is proposed. First, an information transformation mechanism is developed to jointly accommodate three representative heterogeneous preference relations, enabling comparable computations across different preference types. By integrating distances within each preference space into a unified cross-type measurement, a type-robust similarity index is constructed to preserve original semantics and reduce information loss. Simultaneously, building on this metric, an improved similarity-based K-means clustering approach is designed to identify coherent subgroups and derive the weights of decision makers and subgroups. Then, by combining preference similarity with social trust, a fused consensus measure is introduced, together with a personalized consensus feedback strategy that adaptively guides preference adjustments through controllable interventions to efficiently enhance global consensus. A two-stage aggregation and ranking procedure is further employed to obtain the optimal response alternative. Finally, a case study along with sensitivity and comparative analyses demonstrates that the proposed method achieves stable rankings and effective consensus improvement while minimizing information distortion, thereby providing practical decision support for emergency response planning in smart city contexts.

Key words: smart city governance, large-scale group decision making, heterogeneous information, similarity degree, consensus reaching mechanism