Minimum Cost Consensus Model Based on Granular Computing Through Online Reviews for Supporting Group Decision-Making
本文提出一种基于粒度计算和在线评论的最小成本共识模型，通过训练粒度神经网络获取决策者权重和调整成本，并利用评论情感极性动态调整意见，在豆瓣电影数据上验证了方法的灵活性和有效性。
Recently, the minimum cost consensus model (MCCM) has received considerable attention in group decision-making. However, the prevailing trend in early MCCMs is mainly centered on model-centric approaches, which would produce unit adjustment cost and individual options. With the widespread adoption of information technologies and the availability of extensive datasets, accompanied by the advancement of machine learning, constructing MCCMs based on data-driven and machine learning methods, such as neural network (NN), is more practical and reliable. Additionally, considering the inherent variability in models and data, granular computing (GrC) and GrC-based NN can effectively resolve uncertainty. Therefore, this article presents an MCCM based on GrC and online reviews. First, the granular NN (GNN) is trained with online reviews to obtain decision-makers’ weights and unit adjustment costs based on the user information. Subsequently, through assessing the sentiment polarity of decision-makers’ reviews, their opinions are elicited and dynamically adjusted in the form of intervals. Furthermore, the bilateral granular consensus is defined to characterize the impact of decision-makers’ preferences on consensus from both sides. Finally, group decision-making on movies and television productions on douban.com is conducted using the proposed method. Sensitivity analysis demonstrates the flexibility of our consensus approach, and the comparative analysis shows its effectiveness.