基于三层BP网络的多指标综合评估方法及应用

A Method of Multiobjective Synthetic Evaluation Based on Artificial Neural Networks and Its Applications

提出一种基于人工神经网络的综合指标评估方法．文章首先提出了一种新的归一化效用函数，把不同类型，不同量纲的原始评估值转换到［－１，１］区间，该效用函数较好地体现了“奖优罚劣”原则，同时对于神经网络又更容易学习和训练；其次，分析了一般多指标评估中各权值的确定方法及存在的困难；第三，详细介绍了基于人工神经网络的多指标综合评价原理及实现方法，并将之实际应用到一柴油机行业的综合经济效益评估中，取得了较满意的结果．

In this paper, a method of multiobjective synthetic evaluation based on artificial neural networks is presented. First, a new kind of utility function, which changes evaluation values with different dimensions and different kinds of indexes into values in range of with no dimension, not only emphasizes on the principle of 'reward good and penalize bad' but also makes it easy for us to train neural networks to be used in this paper. Second, how to make decision on the weights of indexes is discussed and the difficulties are analyzed. Third, the principle of multiobjective synthetic evaluation based on artificial neural networks is described in detail, and applied to a real synthetic assessment of synthetic benefits of 23 enterprises in China. The results are satisfactory. In the end of this paper, the characteristics of the method proposed in this paper are analyzed.