基于K-means的政府公文聚类方法

Methods of Government Document Clustering Based on K-means Algorithm

政府公文数量巨大,不同政府网站公文分类规则不一,在引用和参考公文时可能发生混淆。针对该问题,基于政府公文题目、摘要和正文内容,采用K-means算法对公文进行分类。首先对政府公文进行分词及去停用词等数据预处理操作,再通过词频-逆文档频率(TF-IDF)权值计算方法,将处理后的政府文本信息转换成二维矩阵,然后采用K-means算法进行聚类。使用清华大学THUCTC文本分类系统对公文聚类结果进行测试。实验结果表明,采用K-means算法对公文进行聚类,准确率达到82.93%,远高于政府网站公文分类准确率。

The number of government documents is huge,and each government body uses different methods to classify their documents,but the results are not satisfactory. The article proposes a method which uses K-means algorithm to classify the government documents based on their titles,abstracts and contents. Firstly,the government documents are preprocessed by word segmentation,removing stop words and etc. Then TF-IDF weight calculation algorithm is used to quantify the government documents. Finally,K-means is utilized to cluster the government documents. THUCTC developed by Tsinghua University is used to test the results of the document classification and the accuracy rate reaches 82.93%,the experiment shows that K-means can achieve higher clustering effects.