| Summary: | 碩士 === 立德管理學院 === 應用資訊研究所 === 91 === Find out frequent itemsets from large data set is a main problem in data mining area, such as the discovery of association rule. We examine the problem of discovering association rules between items in a large database of sales transactions. In this thesis, we propose an efficient Boolean-base algorithm, Boolean Method for Discovering Maximum Frequent Set (BMFS), to discovery the maximum frequent itemset. Base on Pincer-Search algorithm, BMFS use infrequent set to discovering maximum frequent set, and use Boolean algorithm's AND and OR to increase efficiency. This combination leads to three advantages. First, the number of candidate set, BMFS smaller then Boolean algorithm. Second, the efficiency of generate maximum frequent candidate set, BMFS has more efficiency than Pincer-Search algorithm. Third, the number of database scan, in general, BMFS small than Pincer-Search algorithm. We evaluate the performance of algorithm using well-known synthetic benchmark databases. The experimental results show that the BMFS algorithm has better performance than Boolean and Pincer-Search algorithm for most of test cases.
|
|---|
