基于最大频繁Induced子树的GML文档结构聚类-工程技术版.PDF

8 4 () Vo.l 8No. 4 2008 12 JOURNAL OF NAN JING NORMAL UN IVERSITY ( ENG INEERING AND TECHNOLOGY EDIT ION ) Dec, 2008 基于最大频繁 Induced子树的GML文档结构聚类 1 2 朱颖雯, 吉根林 ( 1. , 210012; 2. , 210097) [] Induced GML TBCC lustering. GML Induced, ; CLO E GML , , ; , . TBCClustering, BClustering. [] GML , Induced, Induced [ ] T 391 [] A [ ] 1672-1292( 2008) 04-0050-06 Clustering GML Docum ents by Structure Based on M axim alFrequent Induced Subtrees 1 2 Zhu Y ingwen , JiGenlin ( 1. Departm ent of Computer Elem entary T raining, Sanjiang U niversity, N anjing 210012, China; 2. Schoo l ofM athematics and Computer Science, Nanjing N orm alUniversity, N anjing 210097, China) Abstract: Th is paper presents an algorithm TBCClustering for clustering GM L docum ent structu re based on m ax mi al fre- quent subtree patterns. During them axmi al frequent subtreem ining process, it optmi izes characteristic spaces, gets the m inmi um support automatically, chooses som e subtree pattern to form the optmi istic clustering features, and uses CLO E algorithm to cluster docum ents by clustering features w ithout giving the number of cluster. Not only the dmi ensions of features are reduced, but also the higher clustering precision is obtained. Expermi ent resu lts show thatTBCClustering is m ore effective and efficient than BClustering. K ey words: GML clustering by structure, m axmi al frequent subtrees, closed frequent subtrees