基于云理论神经网络决策树的生成算法.pdfVIP

37 上 海 交 通 大 学 学 报 Vol. 37 Sup. 2003 11 JOURNAL OF SHANGHAI JIAOT ONG UNIVERSITY Nov. 2003 : 1006-2 67( 2003) 1-0113-05 S 田永青, 　杜国宁, 　李　志, 　朱仲英 ( , 200030) : 提出了一种基于五层云神经 络的决策树生成方法. 运用五层云神经 络学习变量间的 云映射关系, 从中生成云决策树. 给出了神经 络的学习算法和云决策树的推理方法, 这种方法不 但具有神经 络的学习能力, 而且结合云理论处理知识的不确定性的能力. 该方法利用了五层云神 经 络学习后的云映射强度, 并能实现云决策树的剪枝优化, 提高了算法的正确率. : 数据挖掘; 决策树; 神经 络; 云模型 : T P 311: A Induct ion of Decision T rees Based on a Cloud Model Neural Network TIA N Yong -qing, DU Guo-ning, LI Zhi, ZH U Zhong-ying ( . , . , 200030, ) Dept of Automation Shanghai Jiaotong Univ Shanghai China Abstract: Cloud mapping relations among variables were learnt by a five layers cloud neural netw ork, then cloud decision trees were extracted from that netw ork. The learning method of neural network and the de- , ducing method of cloud decision tree w ere presented the approach combines the learning ability of neural netw ork w ith the ability of cloud model which deals w ith the ambiguity of know ledge. The approach makes use of the w eights of cloud mappings in the learnt cloud neural netw ork, can realize the optimization of cloud decision trees by branches cutting, and improves the ratio of correctness and efficiency of induction of decision trees. Key words: data mining; decision trees; neural netw ork; cloud model . , 20 60 , . . [ 1] , ( Quilan ID3 ,