透過您的圖書館登入
IP:3.236.240.48
  • 學位論文

模糊關聯規則之研究

A Study on Fuzzy Association Rules

指導教授 : 龐金宗

摘要


隨著計算機技術的迅速發展,其在資料庫中存儲和管理資料的能力正變得越來越重要了。雖然計算機技術的發展有利於數據處理和簡化存儲介質上的要求,提取可用的隱含信息,以幫助決策,已成為一個新的和具有挑戰性的任務。 然而,資料探勘乃是資料的探索和分析,以便於發現有意義的模式。它可以有效地分析各種的應用,這將有助於在業務決策過程。 在資料探勘中,在交易資料庫裡最常見的為關聯規則。其目的在於尋找資料庫裡項目之間的關係,則反映出項目$(X)$出現時,另一個項目$(Y)$可能會出現。 例如,當一位顧客購買麵包時,而也有可能會買牛奶。因此,關聯規則便能協助決策者了解顧客在購買商品時可能的搭配,以便能促進規劃行銷策略。 在本論文,我們結合FP-growth與模糊集合之概念來進行模糊關聯規則探勘。 使用模糊集合是因為它可以自然語言的方式來描述模糊知識,而這也相當符合人們的主觀思考,也更助於增加使用者在制定決策時之彈性。 因此,模糊分割法可模糊的呈現方式是否為使用者所理解。 其次,由於Apriori演算法對於大量資料無法有效率處理的缺點以及時間複雜度會隨著資料愈大而急速的成長,因此在關聯規則的技術上選擇以FP-growth為主。 本論文的目的在於發展處理數量化的模糊資料探勘方法。 基於模糊分割法,由屬性資料中找出可理解與潛在性有用的模糊知識,並進一步結合FP-growth演算法來解決不同的決策問題。 第一個探勘方法是從交易資料中萃取出模糊關聯規則。在此方法中,是利用模糊分割法先轉換成語意的模糊知識,再藉由FP-growth演算法進行探勘。 而在第二個方法中,則是在FP-growth演算法運算的過程中,加入了表格結構來計算,因而能比第一個方法更有效率。 另外,對於需所的參數值及管理者所評估的項目重要性,皆以語意型態來表達,此方式對人類而言是較自然及易於了解。 因此,在第三個方法裡,我們便提出了模糊權重式關聯規則來進行探勘。 最後,本論文所提出的三種不同方式也與其他的方法相比較;其實驗結果也證實本論文所提的方法是具有良好之效率。 除此之外,我們所提的方法皆優於Apriori演算法,而第二個方法則比第一個方法有較佳的效率。

並列摘要


As computer technology progresses rapidly, its capacity to store and manage data in database has become crucial. Though computer technology development facilitates data processing and eases demands on storage media, extraction of available implicit information to aid decision making has turn into a new and challenging task. However, data mining is the exploration and analysis of data in order to discover meaningful patterns. It can effectively be applied on all varieties of analysis and assist the process of decision-making in businesses. In data mining, finding association rules in transaction database is most commonly seen. The purpose is to search for the relation that exists among items of database. The relation reflects that items ($X$) appear, other items ($Y$) are likely to appear as well. For instance, when a customer purchases bread, one might also get milk along with it. Accordingly, association rules can assist decision makers to scoop out the possible items that are likely to be purchased by consumers in the hopes to facilitate marketing strategies. In this dissertation, we combine FP-growth with the concept of fuzzy sets to mine fuzzy association rules. Using fuzzy sets means that we consider fuzzy knowledge representations described by the natural language are well suited for the subjective thinking of human subjects and will assist users in making decisions flexibly. Therefore, the fuzzy partition method can be comprehensible by human users. Next, because Apriori algorithm is not efficient in handing drawbacks for huge data and its time complexity with greater information and rapid growth, the technology of association rule was selected by FP-growth algorithm. The main aim of this dissertation is to develop novel fuzzy data mining techniques for quantitative data to find comprehensible and hidden useful fuzzy knowledge based on the fuzzy partition method and FP-growth to solve various decision problems. In the first algorithm, the fuzzy association rules are extracted from transaction database. In this algorithm, we use fuzzy partition method to transform quantitative data into fuzzy knowledge and apply FP-growth algorithm to mine. In the second algorithm, we increase the table structure in FP-growth algorithm, thus it is more efficient than the first method. In addition, the parameters needed in the mining process and the importance of items evaluated by managers are given as linguistic terms, which are more natural and understandable for human beings. Hence, in the third method, we proposed fuzzy weighted association rule to mine. Finally, we propose three different mining methods to compare with other methods, and the experiment results were verified that the methods proposed have great efficiency. Besides, the proposed approaches are superior to Apriori algorithm. Then the second approach also has more efficiency than the first one.

參考文獻


[7] R. Agrawal, R. Srikant et al., “Fast algorithms for mining association rules,”
[18] Y.-C. Hu, “Mining association rules at a concept hierarchy using fuzzy partition,”
[49] C.-W. Lin, T.-P. Hong, and W.-H. Lu, “An efficient tree-based fuzzy data
quantitative transactions with linguistic minimum supports and confidences,”
[1] G. Piatetsky-Shapiro, “Knowledge discovery in real databases: A report on the

延伸閱讀


國際替代計量