三層式電子目錄之關連規則探勘

Abstract

三層式電子目錄，將電子目錄中的資料描述分為三層，包括了資源目錄階層、類別目錄階層、物品目錄階層。針對三層式電子目錄所儲存的交易資料進行關連式規則的探勘，可提供更詳盡有效的分析資訊。多重階層關連式規則之探勘，主要是利用資料庫中屬性分類架構的特性進行階層式分析，以有效的探勘強關連規則。 本研究以多重階層關連規則探勘方法為基礎，設計在三層式電子目錄中探勘關連規則的方法。所提出的方法，可有效率的發掘單一目錄階層及不同目錄階層間的強關連規則，粹取出隱藏在交易中的資訊，同時利用以分類架構進行階層式探勘的特性，減少候選項目集合之個數，降低掃瞄資料庫之次數。

Three-level E-catalogs store goods information based on three-level metadata including resource level, category level, and goods level. Mining association rules from three-level E-catalogs can provide a detailed and effective way to analyze transaction information. Mining multiple-level association rules uses attribute concept hierarchy to effevtively perform level-wise mining of strong association rules. This study proposes an algorithm of mining association rule from three-level E-catalogs based on techniques for mining multiple-level association rules. The proposed algorithm is able to efficiently discover strong association rules of single level and level-crossing metadata information from three-level E-catalogs. Hidden and detaied association rules can thus be extracted from transaction databases. Moreover the algorithm applies level-wise mining, based on the concept hierarchy, to decrease the number of candidate itemsets and reduce the time of scanning database.