关键词: 贝叶斯网络, 权重, 设备配置

Abstract: A new method was put forward to resolve weights assignment problem of complex system with multi-layer, multi-attribute, multi-object, which firstly used subjective weight to assign attributes, and then adopted Bayes Net to learn the weights. Bayes Net structure was determined according to interrelationships among decision goals. Condition Probability Table was a two-unit framework composed of attribute value and initialized weight. Cupidity Mountain Climbing method was adopted to achieve grads descending strategy in weights self-learning method. Because the Bayes Net had no requirements on the structure of input layer, it could be used to solve weights assignment problem of multi-objective optimization with at least two layers. Weights assigned by this method were obtained through specimen learning by net. Learning process wasnt interfered by jamming, which reduced the influence of uncertain factors in the decision process of complex system so as to guarantee the reliability and veracity of multi-objective decision.

Key words: Bayes Net, weight, equipment allocation, self-learning