CAJ | 학술논문

针对连续属性朴素贝叶斯分类器不能有效利用属性之间的条件依赖信息，而目前所进行的依赖扩展更关注效率，这使得扩展后分类器的分类准确性还有待提高等问题，使用高斯密度估计属性密度，将属性排序、分类准确性标准与属性父节点的贪婪选择结合，综合考虑效率和分类准确性，对朴素贝叶斯分类器进行依赖扩展，建立一阶贝叶斯衍生分类器，并对属性分类提供的信息进行分析。实验结果显示，基于高斯密度的一阶贝叶斯衍生分类器具有良好的分类准确性。
침대련속속성박소패협사분류기불능유효이용속성지간적조건의뢰신식，이목전소진행적의뢰확전경관주효솔，저사득확전후분류기적분류준학성환유대제고등문제，사용고사밀도고계속성밀도，장속성배서、분류준학성표준여속성부절점적탐람선택결합，종합고필효솔화분류준학성，대박소패협사분류기진행의뢰확전，건립일계패협사연생분류기，병대속성분류제공적신식진행분석。실험결과현시，기우고사밀도적일계패협사연생분류기구유량호적분류준학성。
Naive Bayesian classifier with continuous attributes can not effectively use conditional dependency information be-tween attributes.At present,this paper emphasized the efficiency of its dependency extension,which made the classification accuracy of extended classifier need to be improved.Considering the efficiency and classification accuracy comprehensively and basing on Gaussian function to estimate attribute density,the dependency extension of naive Bayesian classifier with con-tinuous attributes was done by combining attribute sorting,classification accuracy criterion and the greedy selection of attribute parent node to establish first-order Bayesian derivative classifier.It also analyzed the information of attributes providing for class.Experiment results show that first-order Bayesian derivative classifiers have very good classification accuracy.