Evaluation of Intelligent Level of Multi-Agent Game Theory Deduction Models Based on Game Neural Networks

The multi-agent game deduction model is a computer-generated and controlled agent that models and simulates human combat behavior. Agents with learning and decision-making capabilities can autonomously respond to events and states in the combat simulation environment, game the war process under the constraints of combat rules, analyze the dynamic evolution of the war process, and deduce the impact of various factors on the outcome of the war. Aiming at the problem of the lack of quantitative evaluation means for the intelligent game theory deduction models, starting from the military needs of joint operation command and decision-making, we constructed a multi-agent game theory deduction model agent level evaluation system, put forward a quantitative method of intelligent level evaluation system for game theory deduction model, and adopted the neural network model of semi-supervised learning to evaluate the model intelligent level, forming the closed-loop of agent training to agent evaluation to agent optimization in game confrontation deduction, through the intelligent level evaluation results of the agent optimization, for the military intelligent game theory deduction research to provide a quantitative assessment of the basis, to enhance the practical value of multi-agent game deduction.