基于改进BIRCH聚类算法的评价对象挖掘  

Opinion Mining Based on BIRCH Clustering Algorithm

作　　者：王梦遥 王晓晔[1,2] 洪睿琪 柴晓瑞 WANG Meng-yao;WANG Xiao-ye;HONG Rui-qi;CHAI Xiao-rui(Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology,Tianjin University of Technology,300191,Tianjin,China;Key Laboratory of Computer Vision and System,Tianjin University of Technology,Ministry of Education,300191,Tianjin,China)

机构地区：[1]天津市智能计算及软件新技术重点实验室,天津理工大学,天津300191 [2]计算机视觉与系统省部共建教育部重点实验室,天津理工大学,天津300191

出　　处：《软件》2019年第11期9-12,61,共5页Software

基　　金：大学生创新训练计划项目(201810060141)的资助

摘　　要：本文对于意见挖掘领域中的评价对象的修剪和聚类问题,提出使用K-means聚类算法和BIRCH聚类算法相结合的方式来进行评价对象的修剪和聚类。利用BIRCH算法类别聚类的功能对评价对象进行聚类,并删除包含较少数据的簇来实现修剪评价对象;再通过对于剩下的簇使用K-means聚类算法来获得最优评价对象。这种修剪聚类方法与以往的基于PMI算法修剪然后基于K-means聚类算法相比,减少了评价对象修剪时对语料库的依赖,最终聚类的结果更加精准,而且BIRCH算法采用一次扫描数据库的策略,可以有效提高速度。For the pruning and clustering evaluation objects in opinion mining, this paper proposes a method that combines BIRCH clustering and K-means clustering algorithm to prune and cluster evaluation objects. Firstly, utilizing BIRCH algorithm of self-learning cluster category, after clustering by BIRCH algorithm, delete the clusters containing few data so that we can prune the evaluation objects. Then use K-means clustering algorithm to make global cluster for the remaining clusters. Compared with pruning using PMI algorithm and clustering using K-means clustering algorithm, our method eliminates the dependency on the corpus. And the cluster result is more accurate. Also BIRCH algorithm scans the database one time, so it can increase the speed greatly.

关 键 词：名词词组模式 BIRCH聚类算法 K-MEANS聚类算法 PMI算法 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]