网页自动分类不确定性问题的贝叶斯网络解法  

Title solution of Bayesian network for uncertain problem of web page classification

作　　者：张晓丹[1] 乔晓东[1] 梁冰[1] 

机构地区：[1]中国科学技术信息研究所,北京100038

出　　处：《计算机工程与设计》2011年第10期3364-3367,3373,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(60803050)

摘　　要：针对网页自动分类中存在的类边界模糊、语料不均匀等引起的分类不确定性问题,提出了贝叶斯网络自动分类融合模型和融合算法,该模型和算法基于网页上多种信息进行融合,并采用不同的与处理方法分别对多种信息进行处理,将处理后的信息输入到贝叶斯网络融合中心进行融合推理,得到最终的分类结果。同时,为了降低贝叶斯网络推理时间复杂度,提出了改进的贝叶斯网络图推理算法。实验结果表明,改进后的融合模型和融合算法能有效解决网页自动分类中的不确定性问题,并能提高网页自动分类的准确率和查全率。For solving uncertain problems of web page classification, a fusion model and fusion algorithm based on multi-kinds information are proposed. The model includes two layers, treatment layer and decision layer. The treatment layer adopts different pre-processing methods for different kind information, the treatment results are input into the decision layer. The final classification result is output from the decision layer after inference. The fusion algorithm adopts the Bayesian network. And an improved Bayesian network inference is proposed for reducing the inference complexity. The experimental result shows that the fusion model and algorithm can solve the uncertain problem of the web page classification and improving classification accuracy and efficiency.

关 键 词：不确定性问题 贝叶斯网络 网页自动分类 贝叶斯推理 信息融合 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]