In group decision-making problems, decision makers prefer to use several linguistic terms to describe their own perception and knowledge, and give their preference intensity of each possible linguistic terms based on their own understanding and interpretation. Due to the nonlinearity of decision maker’s cognition, the gaps between adjacent linguistic terms are unbalanced. The unbalanced probabilistic linguistic term set (UPLTS) is proposed to present such situation. To this phenomenon, a resolution framework is constructed to analyze multiple attribute group decision-making problems under unbalanced probabilistic linguistic environment. Firstly, the integration model based on evidential reasoning theory is proposed to aggregate UPLTSs from different groups in view of incomplete probabilistic distributions in UPLTS. Secondly, the transformation function based on proportional 2 tuple is developed to transform UPLTS into probabilistic linguistic term set, making it easier for subsequent analysis and processing. Thirdly, Based on the multiple types of partitioned structure relationship among attributes, partitioned fuzzy measure is developed to globally capture these interactions among attributes. Then the probabilistic linguistic Choquet integral operator with partitioned fuzzy measure is proposed to obtain the comprehensive performances of alternatives. Lastly, the effectiveness and practicability of the proposed method is demonstrated using three numerical examples and comparing with other methods.

在小组决策问题中，决策者倾向于使用几种语言术语来描述他们自己的感知和知识，并根据他们自己的理解和解释给出他们对每种可能的语言术语的偏好强度。由于决策者认知的非线性，相邻语言术语之间的差距是不平衡的。提出了不平衡的概率语言术语集（UPLTS）来呈现这种情况。针对这种现象，建立了一个解决框架，以分析不平衡概率语言环境下的多属性群决策问题。首先，针对UPLTS中概率分布不完全的问题，提出了基于证据推理理论的集成模型，对不同群体的UPLTS进行汇总。其次，开发了基于比例2元组的转换函数，可将UPLTS转换为概率语言术语集，从而使其易于后续分析和处理。第三，基于属性之间的多种类型的分区结构关系，开发了分区模糊测度以全局地捕获属性之间的这些相互作用。然后，提出了一种基于概率语言的Choquet积分算子，该算子具有划分的模糊测度，以获得替代方案的综合性能。最后，通过三个数值例子并与其他方法进行比较，证明了该方法的有效性和实用性。基于属性之间的多种类型的分区结构关系，开发了分区模糊测度以全局捕获属性之间的这些交互。然后，提出了一种基于概率语言的Choquet积分算子，该算子具有划分的模糊测度，以获得替代方案的综合性能。最后，通过三个数值例子并与其他方法进行比较，证明了该方法的有效性和实用性。基于属性之间的多种类型的分区结构关系，开发了分区模糊测度以全局捕获属性之间的这些交互。然后，提出了一种基于概率语言的Choquet积分算子，该算子具有划分的模糊测度，以获得替代方案的综合性能。最后，通过三个数值例子并与其他方法进行比较，证明了该方法的有效性和实用性。