Secondly, to extract learning samples from the MADM problem, an approach to estimating the utility functions for attributes is presented.

• 其次，提出了基于属性效用函数估计的学习样本构造方法，从决策问题本身抽取学习样本。