基于多分类器融合的防震减灾知识文本分类研究  

TEXT CLASSIFICATION BASED ON MULTIPLE CLASSIFIER FUSION FOR SCIENCE KNOWLEDGE OF REDUCING EARTHQUAKE DISASTER

作　　者：李晓丽 马龙龙 LI Xiaoli;MA Longlong(School of Emergency Management,Institute of Disaster Prevention,Sanhe 065201,China;Institute of Software of Academy of Sciences,Beijing 100190,China)

机构地区：[1]防灾科技学院应急管理学院,河北三河065201 [2]中科院软件研究所,北京100190

出　　处：《高原地震》2020年第3期64-68,共5页Plateau Earthquake Research

摘　　要：防震减灾科普知识文本包括地震监测预报、震害防御和紧急救援等方面的内容。在目前的防震减灾科普知识的宣传应用中,需要人工进行文本类别的选择,面对海量的科普知识文本,人工分类费时费力,自动文本分类是必要选择。针对科普知识文本进行研究,提出了基于多分类器融合的科普知识文本分类方法,通过D-S证据理论融合支持向量机、径向基RBF神经网络和贝叶斯网络三种分类模型获取最终的分类效果。结果表明,通过多分类器融合的分类方法提高了防震减灾科普知识文本分类的性能,其结果明显优于单个分类器。Texts for science knowledge of reducing earthquake disaster include earthquake monitoring and prediction,earthquake prevention and emergency rescue.At present,for the application of popularization of these science knowledge,text classification needs completely manual selection.For large scale texts of science knowledge,artificial classification is a tough work,it is necessary to classify automatically.This paper presents a text classification method based on multiple classifier fusion for science knowledge of reducing earthquake disaster.It is implemented through D-S evidence theory combing support vector machine,radial basis function(RBF)network and Bayesian network classification model.Experimental results show that the performance of text classification in science knowledge of reducing earthquake disaster is improved through multiple classifier fusion method,which is obviously superior to that using single classifier.

关 键 词：文本分类 多分类器融合 防震减灾 多数投票法 

分 类 号：P315-39[天文地球—地震学]