結合約略集理論與關聯法則於順序資料分析之研究

機構典藏 > College of Business and Management > Department of Management Sciences > Thesis > Item 987654321/77385

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/77385

Title:	結合約略集理論與關聯法則於順序資料分析之研究
Other Titles:	The study of integration of rough set theory and association rules for ordinal data analysis
Authors:	陳盈如;Chen, Yin-Ju
Contributors:	淡江大學管理科學學系博士班廖述賢
Keywords:	約略集理論;資料採礦;關聯法則;Rough set theory;data mining;association rule
Date:	2012
Issue Date:	2012-06-21 06:38:43 (UTC+8)
Abstract:	首先，傳統的關聯法則，使用者必須不斷的試誤（包含：屬性的挑選、門檻值的設定等…法則產生前的相關程序與步驟），俾便找出具解釋能力的關聯法則。再者，與近期相關研究相比，資料採礦資料都是以資料是精確且乾淨為前提的，在這樣的條件下所產生的關聯法則，可能會發生在某些特定情況下（例如：有人為輸入的錯誤、記錄錯誤等…不完整資料），符合條件的規則被淘汰亦或產生過多的規則。最後，透過相關研究的文獻探討，發現約略集理論已成功的被運用在選擇屬性及改變效率之決策問題上。因此，本研究選擇以約略集理論為研究的理論基礎，從縮短決策者探勘關聯法則的試誤時間為解決問題的方向，在規則產生前，利用集合的產生，針對資料型態涉及順序尺度或含區間資料的順序尺度，提供新的演算概念。希冀，在不失去原本的排序關係的前提下，提供更多的排序資訊予決策者使用。研究中，針對順序尺度與含區間資料的順序尺度，分別提出約略關聯法則的探勘步驟、演算流程說明、應用於酒精飲料產品與非酒精飲料的案例，以及提供相關個案的管理意涵。最後，將本研究所未考量到的部分以及可以持續研究的方向分段論述，讓後續的相關研究學者可以參考。 First, as per the traditional association rules, in order to identify meaningful association rules, the user must use trial and error method （including attribute choice, threshold value hypothesis, etc., considering the procedure and step taken before the association rules were formulated）. Furthermore, unlike algorithm-related research, data mining algorithms assumed that input data were accurate; however, the assumption would not be made in case one best rule exists for each particular situation such as input mistake or record mistake and similar incomplete data. Finally, through literature review, rough set theory has been successfully applied in deriving decision trees/rules and specifying problems, with proven effectiveness in selecting attributes. Therefore, we select rough set theory on the basis of our research, and this reduces the time that policymakers take to determine meaningful association rules. Before the rule is formulated, through the set process, we provide a new algorithm for the data type that involves ordinal data and ordinal data with internal data. Under a condition that does not affect the sorting relations between the values of the ordinal data, we provide more sorting information that the policymakers can use. In the research, we provide two new algorithms that are suitable for ordinal data and ordinal data with internal data. Further, we provide illustrative examples using alcoholic and non-alcoholic beverage products individually. Finally, we give some suggestions for future research.
Appears in Collections:	[Department of Management Sciences] Thesis

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	320	View/Open

Loading...