一种基于变调整学习规则的模糊网页分类方法研究

A Fuzzy Classification of Web Pages Based on the Transposition-Learning Rule

当类别之间交叉现象比较严重时,网页分类方法的精度就会下降.为准确地分类网页,首先给出一种模糊网页分类的系统结构,通过用成员函数替代分类网络中的权值变量,来提供一种可融入人类关于网页分类知识的机制.然后给出一种通用学习规则,来学习成员函数中的参数.通过理论推导,用李雅普诺夫函数分析和验证通用参数学习规则的学习收敛性,揭示参数学习算法朝最小误差方向调整参数的内在因素.最后在单参数学习算法收敛性的分析基础上,提出一种变调整规则的单参数学习算法,加快参数学习速度.从学习收敛性的理论论证和实验结果来看,这种网页分类方法是一种有效的分类方法.

When the overlap of categories is excessive, the accuracy of Web page classification decreases. In order to classify the Web pages accurately, a framework of fuzzy classification of Web pages is presented, to give a mechanism of combining the human knowledge on Web page classification by a member function. Then a general learning rule of the coefficients is proposed. The Lyapunov function is used to analyze the convergence of the general learning rule, and it is proved in theory that the general learning rule has the inherent factor which adjusts the coefficient values to gain the minimum error. On the basis of theoretic convergence analyses of a single-coefficient learning algorithm, a transposition rule is proposed, which is applied to the single-coefficient learning algorithm to gain quick convergence speed in the phase of coefficient learning. It is shown that both from the theoretic deduction of the learning convergence and from the experiment result, the fuzzy classification of Web pages is an efficient method.