A Novel Self-Adaptive Bayesian Belief Network with Information Entropy for Leveraging Decision Consensus among Multi-Agents

Decision making in a semistructured or unstructured problem should consist of a combination of solution procedures, probabilistic reasoning, and human judgments. Such processes mostly involve in evaluating multi-objectives or attributes while making decisions. In addition, they are always complied with the decision makers’ inward probabilistic structure, which will be tedious without the assistance of Decision Support System. In this study, a novel mechanism for Multi-intelligent Agents to reflect probabilistic reasoning from domain experts, negotiate among decision groups, share mutual information, and evaluate multi-decision variables to generate corresponding unique solution was proposed. The decision makers’ reasoning experience and styles are configured by the Interior Bayesian Belief Network (BBN), pooling factors, and adaptation parameters while employing Intelligent Agents to negotiate and evaluate events coordinately. The negotiation mechanism is based on information theoretic measurements to automatically pool group members’ opinions and then trigger self-adaptations of agents. Exterior BBN also exhibits the dependence of group members in order to model the mutual impacts among them. Finally, an illustrative example was presented, in which the characteristics of the proposed system was clarified.