This paper investigates the diversified multi-attribute group decision-making problem which means that the different decision-makers have different evaluation attribute sets for all candidate alternatives. We analyze the diversified multi-attribute group decision-making problem from the perspective of granular computing models and creatively combine soft sets with multigranulation fuzzy rough sets to construct the multigranulation soft fuzzy rough set model. Subsequently, we define the optimistic upper and lower approximations and pessimistic upper and lower approximations with respect to multigranulation soft fuzzy rough set model. Meanwhile, we discuss some important mathematical conclusions and properties based on the established model. Then, we propose a novel approach to the diversified multi-attribute group decision-making problem based on multigranulation soft fuzzy rough set and TODIM method, which fully consider the reference dependence and loss aversion of decision-makers in the decision-making process. Finally, we use a numerical example of evaluation for the policy selection group decision-making problem to describe and explain the calculation principle and process and compare the results with existing related work, and our approach has great stability and rationality. And the contributions of this paper include the following two points: (1) construct the multigranulation soft fuzzy rough set model; (2) combine the established model with TODIM method to provide a novel perspective to solve the diversified multi-attribute group decision-making problem.

本文研究了多元化的多属性群体决策问题，这意味着不同的决策者对所有候选方案具有不同的评估属性集。我们从粒度计算模型的角度分析了多元化的多属性群决策问题，并将软集与多粒度模糊粗糙集创造性地结合起来，构建了多粒度软模糊粗糙集模型。随后，对于多粒度软模糊粗糙集模型，我们定义了乐观的上下近似和悲观的上下近似。同时，我们基于建立的模型讨论了一些重要的数学结论和性质。然后，我们提出了一种基于多粒度软模糊粗糙集和TODIM方法的多元化多属性群决策方法，该方法充分考虑了决策者的参考依赖性和决策者的损失厌恶。最后，通过一个数值算例对政策选择群体决策问题进行了描述和解释，并说明了计算原理和过程，并将结果与​​现有相关工作进行了比较，方法具有较强的稳定性和合理性。论文的主要工作包括以下两点：（1）构建了多粒度软模糊粗糙集模型；（2）将建立的模型与TODIM方法相结合，为解决多元化的多属性群体决策问题提供了新的视角。