从RBF核函数中抽取关联分类规则

目的为解决SVM分类器的分类模式难以为人类专家所理解等问题而提出一种有关InterRBF算法的新思路。方法通过将RBF核函数将其展开成麦克劳林级数,并从展开式中挖掘对分类分析起重要作用的关联规则,从而在SVM的分类模式中学习出关联规则分类器。结果改进后的SVM分类器具有较好的分类准确度;改变了当前研究从SVM的分类模式中抽取规则的方法仅限于IF-TEHN规则或者学习出决策树的状况。结论从RBF核函数抽取关联分类规则,对于在难以理解的知识中提取可理解的表达规则是可行的方法。

Extracting association classification rules from RBF kernel

Aim To solve the problem that the classification model of SVM classifiers is non-understandable to human experts,Inter RBF algorithm was presented.Methods The method expands RBF kernel into its Maclaurin series,and mines association rules which make great contribution to classification from this series,so as to learn association classifier from the SVM classification model.Results SVM classifiers modified have excellent classification accuracy.The research changes the status that current methods to extract rules from SVM classification model are limited in extracting IF-TEHN rules or learning decision trees.Conclusion RBF kernel function associated classification rules,which are taken from this approach for extracting the knowledge to understand the rules of understandable expression,are feasible.