| 作成者 |
|
|
|
|
|
|
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 出版タイプ |
|
| アクセス権 |
|
| 関連DOI |
|
|
|
| 関連URI |
|
|
|
| 関連情報 |
|
|
|
| 概要 |
We address the problem of finding interesting substructures from a colletion of semi-structured data such as XML or HTML. Our framework of data mining is optimized pattern discovery introduced by Fuku...da et al., where the goal of a mining algorithm is to discover a pattern that optimizes a given statistical measure such as the information entropy over a class of simple patterns. In this paper, modeling semi-structured data with labeled ordered trees, we study the efficient algorithm for the optimized pattern discovery problem for the class. In a previous paper, we developed the rightmost expansion technique and the incremental occurrence update technique by generalizing enumeration technique developed by Bayardo (SIGMOD'98) for discovering long itemsets to implement an efficient frequent pattern miner for the class of labeled ordered trees. By combining these technique with the pruning technique for optimized patterns of Morishita and Sese (PODS'00), we present an efficient algorithm for finding optimized patterns for labeled ordered trees of bounded size. Experimental results show that our algorithm perform well on a variety of size of data and range of parameters. We also show an approximation hardness result for labeled ordered trees of unbounded size.続きを見る
|
Mendeley出力
trcs206
trcs206.ps