Abstract: To solve the problem of getting a more precise result of comprehensive evaluations, fuzzy neural network method is introduced. This method uses a network training algorithm to cope with subjective factors in evaluations model. The algorithm can efficiently approximate the weight set and degree of membership function. To enhance its validity, many skills are adopted to optimize the network training process. The results show that the strategy can efficiently improve the final precisions. It is a better solution for synthesis evaluations problem.

Key words: comprehensive evaluations, neural network, weight, evaluation matrix

摘要： 针对综合评判模型中主观因素影响评判精度问题，引入模糊神经网络优化评判模型的方法，采用模糊神经网络的训练过程逼近综合评判模型的权重集和隶属度函数。为了提高该方法的有效性，对网络训练算法进行多方面的优化，以便神经网络更快更稳定地收敛。结果证明该方法能较好地去除综合评判模型中的主观成分，有效地提高评判的最终精度。

关键词: 综合评判, 神经网络, 权重, 评价矩阵

CLC Number: 

• TP183