网络舆情倾向性预测模型仿真研究  

Simulation Research of Network Public Opinion Orientation Prediction Model

作　　者：连淑娟[1] 薛玉[1] 

机构地区：[1]北京外国语大学信息技术中心,北京100089

出　　处：《计算机仿真》2016年第3期323-326,338,共5页Computer Simulation

基　　金：中央高校基本科研业务费专项资金资助(2013JJ006)

摘　　要：在对网络舆情倾向性预测建模过程中,由于网络舆情倾向性话题评论频繁,数目众多,使得网络舆情倾向性受到特征词语出现频率的影响产生无规律性。传统的网络舆情倾向性预测方法,主要是以特征词语出现频率为前提来建模,忽略了少数词语会对舆情倾向性产生翻转性影响因素,导致网络舆情倾向性预测分析准确性差的问题。提出文本网络舆情倾向性预测模型,将中文自动构词算法的SVM文本倾向性分类方法词库和信息瓶颈的特征提取方法相结合,自动化的动态构建专用词库,引入信息瓶颈概念,根据文章关键词的性质进行分类,在结合概念抽取的方法,将词聚类结果映射到知网义原,并作为分类特征完成网络舆情倾向性的预测。仿真结果表明,上述方法能够有效提高网络舆情倾向性的预测准确度。Due to its frequent and numerous, the network public opinion orientation is affected by the frequency characteristics of words with no regularity. A prediction model for network public opinion orientation is put forward based on the text, the classified word library of SVM text orientation of Chinese automatic word formation algorithm is combined with the information bottleneck feature extraction method, and the special word library is built dynamically and automatically. A concept of information bottleneck is introduced, and it is classified depending on the nature of the keywords. Combined with the concept extraction, the word clustering result is mapped to the hownet original right- eousness, and is taken as classification characteristics to complete the prediction of network public opinion orienta- tion. The simulation results show that the method can effectively improve the prediction accuracy of network public o- pinion orientation.

关 键 词：自动构词算法 支持向量机文本 信息瓶颈 特征提取 关键词聚类 概念抽取 

分 类 号：TP133[自动化与计算机技术—控制理论与控制工程]