A Two-Stage Method for Improving the Decision Quality of Consensus-Driven Three-Way Group Decision-Making

In three-way group decision-making based on the minimum risk Bayesian decision theory, the consensus of basic loss function evaluation becomes its main core issue. However, if we only consider evaluation information consensus, it does not ensure the classification quality of three-way decisions. Thus, to balance the consensus and decision quality, we design a three-way group decision-making joint learning process via constructing a two-stage group consensus method. Inspired by supervised learning, Stage 1 establishes the minimum decision error optimization model (MDEOM) to learn the optimal parameters of three-way decisions and calculate decision loss reference values. Then, we design an algorithm to solve MDEOM based on particle swarm optimization (PSO) algorithm. In Stage 2, we calculate adjusted decision losses with the minimum decision loss difference consensus model (MDLDCM), which can guide the consensus adjustment of loss functions and improve the decision quality of three-way group decision making. Finally, some implications of three-way group decision making with the two-stage group consensus method are discussed by a series of experiments which prove the improvement effectiveness of our proposed method.