A novel heterogeneous preference fusion method with dual interaction influence and dynamic risk spread

Abstract

Preference fusion is crucial for multi-attribute group decision-making (MAGDM). Heterogeneous preferences and interactions are commonplace situations in real-world MAGDM. Unlike conventional MAGDM fusion methods ignoring interactions or only considering attribute interactions, a novel heterogeneous preference fusion method proposed in this study considers the dual interaction influence of attributes and alternatives, as well as the influence of the dynamic risk spread (IDRS). Specifically, a novel similarity-based three-dimensional determination–uncertainty space method is established to homogenize heterogeneous preferences, avoiding information distortion caused by preference uncertainty loss. The attribute interaction influence is evaluated by constructing an optimization model with ranking maximization. Subsequently, a multi-attribute alternative interaction network and a total interaction measurement model are developed to describe complex interaction relationships and the interaction intensity between alternatives. The susceptible-infected-recovered-susceptible Markov chain model is presented to measure the IDRS. An improved directed weighted degree centrality model based on the IDRS is then established to measure the alternative interaction influence (AII) from a global perspective. Finally, a novel interaction non-additive fusion method is designed to fuse preference more reasonably. A case study of hot dry rock siting and comparative analyses demonstrate the practicality, robustness, and superiority of the proposed method. The results show that the AII and IDRS significantly influence the fusion accuracy and should be considered in MAGDM fusion methods.