A personalized individual semantics-based multi-criteria method for learning preferences in the graph model for conflict resolution
提出一种基于个性化个体语义的多准则评价方法，通过异构语言评价矩阵和偏好示例学习决策者偏好，用于冲突解决图模型中的稳定性分析。
In the graph model for conflict resolution (GMCR), accurately determining the preferences of decision makers (DMs) over states is crucial. DMs’ preferences are often influenced by multiple factors (criteria), which can be elicited using a multi-criteria evaluation method. Moreover, in practice, DMs prefer to utilise linguistic information to express preferences, potentially employing different linguistic evaluation formats across criteria. Given that words may convey different meanings to different individuals, DMs often exhibit personalised individual semantics (PIS) in their linguistic preferences. Therefore, a PIS-based multi-criteria evaluation method is proposed to learn preferences within GMCR. In this method, DMs express preferences using heterogeneous linguistic evaluation matrices (HLEMs) and provide preference examples in the form of relative preference relations between states. A deviation minimum-based optimisation model is designed to personalise individual semantics by minimising the deviation between preference examples and relative preference values derived from HLEMs. Then, a consistency improvement model is developed to improve the consistency between the two types of preference information by removing the minimum number of inconsistent preference examples. This is followed by obtaining the comprehensive evaluation value vectors of states from HLEMs for stability analysis. Finally, the Elmira conflict is utilised to demonstrate the application of the proposed method.