決策邏輯型機制及其在知識表徵中之應用

Abstract

近年來，從資料庫中發掘知識與其核心機制—資料探勘越來越受到廣泛的注意。雖然資料探勘研究一貫是以設計高效率的演算法為主，然如何使探勘所得的知識能對使用者有用，仍然持續成為該領域一個最具挑戰性的問題。由於知識能對使用者有用的先決條件是使用者能瞭解其意義，因此知識表徵機制在知識管理過程中便扮演著重要的角色。 我們在本學位論文中探討的就是從粗略集合論觀點對決策邏輯作若干擴充。傳統決策邏輯是以粗略集合論為基礎的資料探勘中一種標準的知識表徵機制，而我們的擴充顯示出決策邏輯型機制對於較複雜的知識管理工作亦非常有用。 我們一方面提出數種決策邏輯語言，可用於表達以粗略集合論為基礎的多準則決策分析方法所產生的決策法則，這些語言的語意模型為表示多準則決策記錄之資料表，其中每一個決策記錄都可以用有限多個屬性或準則來描述。而準則與屬性的差別是屬性值之間不見得有優劣關係，而準則的值之間必然存在優劣關係。 另一方面，我們提出矢決策邏輯來對關聯資訊系統中所發掘出來的知識作表徵與推理。此一邏輯係結合矢邏輯與決策邏輯的主要特徵而成，其中矢邏輯為一種用於關係推理的樣態邏輯。矢決策邏輯的邏輯式可以在關聯資訊系統中加以解釋，而關聯資訊系統不僅描述物件的屬性，亦描述其彼此之間的關係。我們提出一種矢決策邏輯的公設化系統，證明其完備性，並展現其在多準則決策分析與社交網路分析上的應用。

我們的結果對知識管理中知識表徵此一環節特別有用。我們以一個現實的例子來說明我們所提出來的機制可用來輔助公司聘僱人員及形成團隊過程中不同階段的知識表徵需求。

In recent years, knowledge discovery in databases (KDD) and its kernel data mining have received more and more attention for practical applications. While the mainstream research of data mining concentrates on the design of efficient algorithms for extracting knowledge from databases, the question to close the semantic gap between structured data and human-comprehensible concepts has been a lasting challenge for the research community. Since the discovered knowledge is useful for a human user only when he can understand its meaning, the representation formalism will play an important role during the knowledge management life cycle. In this dissertation, we investigate several extensions of decision logic (DL) from the perspective of rough set theory. Traditionally, DL has been considered as a standard way of knowledge representation for rough set-based data mining, whereas our extensions show that DL-styled logics are also useful in more complicated knowledge management tasks. On the one hand, we propose some decision logic languages for rule representation in rough set-based multicriteria decision analysis. The semantic models of these logics are data tables representing multicriteria decision records. Each decision record is described by a finite set of criteria/attributes. The domains of the criteria may have ordinal properties expressing preference scales, while the domains of the attributes may not. On the other hand, we propose an arrow decision logic (ADL) to represent and reason about knowledge discovered from relational information systems (RIS). The logic combines the main features of decision logic (DL) and arrow logic (AL). AL is the basic modal logic of arrows. ADL formulas are interpreted in RIS which not only specifies the properties of objects, but also the relationships between objects. We present a complete axiomatization of ADL and discuss its application to knowledge representation in multicriteria decision analysis and social network analysis. Our work is particularly useful for the knowledge representation phase in the knowledge management life cycle. A realistic scenario about human resource management is used to show how the proposed logics can serve as representational formalisms in different stages of the recruitment process and team formation process of a company.