An adaptive algorithm for consensus improving in group decision making based on reinforcement learning

In group decision-making problems with reciprocal preference relations, the process of improving individual consistency and consensus degree among decision-makers is dynamic and iter ative. Traditional automatic consensus-reaching process has some problems, such as adopting a fixed strategy in a deterministic environment without considering the dynamics of the decision-making environment, the destruction of individual consistency, etc. To solve these problems, an adaptive consensus-reaching model in a dynamic environment is proposed in this paper. Firstly, a Q-learning algorithm is used to build an environment model for different decision-making states of matrix modification. On the premise of modifying the preference matrix with a small matrix deviation, the optimal modification strategy is learned to improve the consensus degree among decision-makers. Second, we propose a method to control individual consistency in the process of consensus reaching by using a reward function. Finally, several numerical examples are used to illustrate the effectiveness and feasibility of the proposed algorithm. The experimental results show that the proposed algorithm significantly improves the consensus degree of the decision-makers in a small matrix deviation and ensures that the decision-maker’s individual consistency is not destroyed.