預測式關聯規則演算法

CCUR > College of Business > Department of Information Management & Graduate Institute of Information Management > Thesis > Item 987654321/20324

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/20324

Title:	預測式關聯規則演算法
Authors:	李勃穎
Contributors:	資訊管理學系
Keywords:	資料探勘關聯規則演算法 apriori演算法
Date:	2010
Issue Date:	2011-11-14 12:43:15 (UTC+8)
Abstract:	近年來關聯規則(association rules)技術已被廣泛的運用在資料探勘領域之中，關聯規則演算法分為兩個部份，首先從交易資料中找出購買次數高於支持度門檻的頻繁項目集，其次為這些頻繁項目集中找出商品之間購買的規則。在關聯規則演算法中，時間的耗費主要在於找到頻繁項目集，而在以往的關聯規則演算法之中最常被使用的為Apriori演算法，雖然此演算法可以找出頻繁項目集，但是它存在著兩大缺點，第一點為產生過多的候選項目集，第二點為需多次掃瞄資料庫，而造成整體執行時間效率不佳。許多專家學者針對這兩個缺點提出改善的方式，然而皆未離開Apriori的架構，因此本論文提出預測式關聯規則演算法來提昇找到頻繁項目集的時間效率。在預測式關聯規則演算法中，只需掃瞄資料庫兩次，第一次掃瞄完成長度項目分配表，接下來再利用使用者所輸入的長度誤差容許和頻繁誤差容許預測出所有的頻繁項目集，接著再一次掃瞄資料庫找出頻繁項目集，其優點為執行時間效率佳，缺點為可能產生誤差。 In past decades, the association rules technology has be applied in data mining domain. The association rules algorithm has two parts. The first part is finding the frequent item set where purchase of times over support threshold from transaction data. The second part is finding the association rules from frequent item set. In the association rules algorithm, the first part is time-consuming. Apriori algorithm is the most often used association rules algorithm in former association rules algorithm. Although, this algorithm can finding the frequent item set. But it has two shortcomings. The first shortcoming is generating candidate item set too much. The second shortcoming is scanning transaction data times without number. Therefore give occasion to time-consuming. Many experts propose the improvement ways in view of these two shortcomings. However the improvement ways are still using Apriori algorithm structure. In this paper we propose predictive association rules algorithm. This algorithm can finding the frequent item set quickly. Predictive association rules algorithm only need scan the database two times. First scan finish length-Item distribute total table. Then using the length error number and frequent error number predictive all frequent item set. Finally scan the transaction finding real frequent item set.
Appears in Collections:	[Department of Information Management & Graduate Institute of Information Management] Thesis

Files in This Item:

File	Size	Format
224封面.pdf	98Kb	Adobe PDF	321	View/Open
224摘要謝辭目次.pdf	205Kb	Adobe PDF	824	View/Open
224全文.pdf	1134Kb	Adobe PDF	2524	View/Open

Loading...