| Abstract: | 在真實世界中，交易資料通常包含數值型資料。因此，許多利用模糊理論的方法被提出用來從數值型資料中探勘模糊關聯規則。此外，由於每個商品有它自己的利潤，最近這幾年利潤商品集探勘領域也變得相當受歡迎。然而這些方法的共通問題，第一、這些方法的最小支持度不易設定；第二、所探勘出的規則只揭露常識性的資訊，使得這些規則不具有商業價值。故本論文中，我們提出兩個具有命題邏輯特性的演算法，分別為模糊一致性規則(FCR)及高一致性利潤模糊商品集(HCUFI)去克服上述問題。
在第一個演算法，它首先將數值型資料轉換成模糊集合。之後，根據產生出來的模糊集進一步產生候選模糊一致性規則。最後，計算每條候選模糊一致性規則的列聯表，並利用它來檢查規則是否有滿足命題邏輯的四個標準。如果是，該規則就是一條模糊一致性規則。
而在第二個演算法中，由於商品有不同的利潤，因此，我們首先提出一個領域衍生模糊資探勘架構，根據此架構，我們進一步提出一個高一致性利潤模糊商品集探勘演算法來提升探勘樣式的商業價值。它首先轉換數值型交易資料成模糊集合。然後，利用外部利潤表計算每個模糊商品集利潤值，假如此利潤值大於或等於最小利潤比率值，這個商品集會被當作高利潤模糊商品集。最後，計算高利潤模糊商品集的列聯表，並檢查它是否有滿足命題邏輯的四個標準。如果是，該商品集就是一個高一致性利潤模糊商品集。
在實驗部份，透過foodmart資料集亦顯示所提的兩個演算法是有效的。因此，第一個演算法的優點是在不用設定最小支持度下可以自動探勘具有符合命題邏輯且具商業價值的規則。而第二個演算法的優點則是透過領域衍生探勘架構，所提的方法可以探勘出更貼近商業需求的可行動知識樣式。
In real-world applications, transactions usually contain quantitative values. Many fuzzy data mining approaches have been proposed for finding fuzzy association rules from the give quantitative transactions. In addition, since each item has its own utility, utility itemset mining has thus become an interesting field in recent years. However, the common problems of those approaches are that: first, an appropriate minimum support is not easy to be set; second, the derived rules usually expose common-sense knowledge which may not interesting in business point of view. In this thesis, we thus propose two algorithms, called Fuzzy Coherent Rules (FCR) mining and High Coherent Utility Fuzzy Itemsets (HCUFI) mining, to overcome the mentioned problems with the properties of propositional logic.
The first algorithm first transforms quantitative transactions into fuzzy sets. Then, those generated fuzzy sets are further collected to generate candidate fuzzy coherent rules. Finally, contingency tables for every candidate fuzzy coherent rules are calculated and used for checking those candidate fuzzy coherent rules satisfy four criteria or not. If yes, it is then a fuzzy coherent rule.
In second algorithm, due to each item has its utility, we first propose a domain-driven fuzzy data-mining framework. According to the framework, we further propose a high coherent utility fuzzy itemsets mining algorithm for increasing patterns’ business merits. It first transforms quantitative transactions into fuzzy sets. Then, utility of each fuzzy itemsets is then calculated according to the given external utility table. If the value is large than or equals to the minimum utility ratio, it is considered as high utility fuzzy itemset (HUFI). Finally, contingency tables are calculated and used for checking those HUFIs satisfy specific four criteria or not. If yes, it is a High Coherent Utility Fuzzy Itemsets (HCUFI).
Experiments on the foodmart dataset have also been made to show the efficiency of these two proposed approaches. The advantage of first algorithm is that it can derive business interestingness rules with propositional logic without setting minimum support. And, the advantage of second algorithm is that it can derive more actionable knowledge pattern with business interestingness based on the domain-driven fuzzy data-mining framework. |
