基于Bagging的XML文档集成聚类研究

Study of XML documents ensemble clustering based on Bagging

将集成学习方法应用到XML文档聚类中来改进传统聚类算法的不足。提出一种标签与路径相结合的XML文档向量模型,基于这个模型,首先对原始文档集进行多次抽样,在新文档集上进行K均值聚类,然后对得到的聚类中心集合进行层次聚类。在人工数据集和真实数据集上的实验表明,该算法在召回率和精确率上优于K均值算法,并且增强了其鲁棒性。

A method of ensemble learning is applied in XML documents clustering in order to improve the clustering performance.A novel vector model based on tag-path of XML documents is proposed and the documents are mapped to the model.The original datasets is sampled into several Bootstrap datasets,K-means algorithm is first run on each of the Bootstrap datasets,then hierarchical clustering algorithm is run on the sets of K-means clusters centers.The experimental result on the synthetic and real datasets shows that this algorithm is superior to the K-means algorithm on recall rate and precision rate,and enhances the robust of K-means algorithm.