Coming to Consensus on Classification in Flexible Linguistic Preference Relations: The Role of Personalized Individual Semantics

In group decision making problems, linguistic approaches are extensively utilized to convey the preferences of decision makers. It is common for words to mean different things to different people, thereby requiring the modeling of the personalized individual semantics (PIS) of decision makers when employing linguistic approaches. This paper investigates the PIS-based ordinal classification group decision making problem, in which flexible linguistic preference relations are used to convey decision makers’ preferences. Firstly, a minimum information violation model with PIS is proposed to address the PIS issue in flexible linguistic preference relations, while simultaneously obtaining the ordinal classification of alternatives. When the information violation level derived from the minimum information violation model with PIS is unacceptable, a minimum adjustment-based ordinal classification consensus model is presented to obtain the references for decision makers to modify their preferences. Subsequently, an interactive ordinal classification consensus reaching process is devised, which aims to help decision makers to reach the predefined information violation level. Finally, to justify the validity of the proposal, a numerical example regarding research and development project selection, a comparative analysis, and a sensitivity analysis are provided.