欢迎来到天天文库
浏览记录
ID:33971874
大小:616.26 KB
页数:5页
时间:2019-03-02
《基于机器学习的中文微博情感分类实证研究》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/321588841ThearticletitleArticleinComputDes·December2017CITATIONSREADS06921author:SeanYoungChineseAcademyofSciences18PUBLICATIONS0CITATIONSSEEPROFILESomeoftheauthorsofthispublicationarealsoworkingo
2、ntheserelatedprojects:ThisisthirdprjectViewprojectasdffsdfsdaasfdafViewprojectAllcontentfollowingthispagewasuploadedbySeanYoungon06December2017.Theuserhasrequestedenhancementofthedownloadedfile.ComputerEngineeringandApplications计算机工程与应用2012,48(1)1⦾博士论坛⦾基于机器学习的中文微博情感分类实证研究刘志明,刘鲁LIUZhiming,LIULu北京
3、航空航天大学经济管理学院,北京100191SchoolofEconomicsandManagement,BeihangUniversity,Beijing100191,ChinaLIUZhiming,LIULu.EmpiricalstudyofsentimentclassificationforChinesemicroblogbasedonmachinelearning.ComputerEngineeringandApplications,2012,48(1):1-4.Abstract:Withthedevelopmentofmicroblog,itismoreconvenientto
4、commentontheWeb.Uptonow,thereareveryfewstudiesonthesentimentclassificationforChinesemicroblog,thereforethispaperusesthreemachinelearningalgorithms,threekindsoffeaturese-lectionmethodsandthreefeatureweightmethodstostudythesentimentclassificationforChinesemicroblog.Theexperimentalresultsindicateth
5、attheperformanceofSVMisbestinthreemachinelearningalgorithms,IGisthebetterfeatureselectionmethodcomparedtotheothermethods,andTF-IDFisbestfitforthesentimentclassificationinChinesemicroblog.Combiningthethreefactorsthecon-clusioncanbedrawnthattheperformanceofcombinationofSVM,IGandTF-IDFisbest.Forthe
6、moviedomainitisfoundthatthesentimentclassificationdependsonthereviewstyle.Keywords:microblog;sentimentclassification;machinelearning;featureselection;termweight摘要:使用三种机器学习算法、三种特征选取算法以及三种特征项权重计算方法对微博进行了情感分类的实证研究。实验结果表明,针对不同的特征权重计算方法,支持向量机(SVM)和贝叶斯分类算法(NaïveBayes)各有优势,信息增益(IG)特征选取方法相比于其他的方法效果明显要好。
7、综合考虑三种因素,采用SVM和IG,以及TF-IDF(TermFrequency-InverseDocumentFrequency)作为特征项权重,三者结合对微博的情感分类效果最好。针对电影领域,比较了微博评论和普通评论之间分类模型的通用性,实验结果表明情感分类性能依赖于评论的风格。关键词:微博;情感分类;机器学习;特征选取;特征项权重DOI:10.3778/j.issn.1002-8331.2012.01.001文章编号:1002-8331(
此文档下载收益归作者所有