基于Rough集的规则抽取技术

数据分类是数据挖掘的一个重要功能,神经网络以其良好的抗噪性和鲁棒性而成为一种广泛使用的数据挖掘工具,尤其是运用在数据分类中.但是,神经网络对用户来说是一个黑箱,所获得的知识隐含在神经网络的连接权中而难以理解.针对这种情况,建立了一个基于神经网络的数据分类系统模型,通过数据处理、网络训练、规则抽取等几个阶段,达到将获得的知识清晰化的目的.在系统中,首先对连续性数据作规一化和对语义性数据进行编码;然后经过网络训练而获取知识;规则抽取采用功能性方法:即把神经网络视为黑盒,随机产生输入得到相应的输出组成实例,然后采用Rough集的方法进行约简得出规则.

Extracting Rule from the Neural Network Based on Rough Set

Data classification is the important function of the data mining and neural network which has anti-noise and robust becomes a data mining tool and especially in the data classification.But,neural network is a black-box for the user and for its knowledge implies the weight of the conjunction,the user can not comprehend the knowledge.In order to resolve this problem,in this paper,we build a data classification model based on the neural network.This data classification can make the knowledge understood by using the data process,training neural network and extracting rule.In this data classification,first,we generalize the continued data and encode the nominal data.Second,we acquire the knowledge by training the neural network.Last,we extract the rule from the neural network by the function method.In this method the neural network is looked as a black-box and we input a random input-data and then the network can create the output.An instance consists of the random input-data and output.We reduce the instances and can acquire the rules by using the rough set.