| Summary: | 碩士 === 大葉大學 === 電機工程研究所 === 90 === Data mining is a very important database research issue. Especially, the generation of frequent patterns in large database has been widely studied. Most of the studies take the Apriori-based approach, which has great effort in the generation of candidate frequent patterns and needs multiple database accesses. The FP-tree-based approaches have been proposed to avoid the generation of candidate set and scan transaction database only twice, but they work with more complicated data structure. Recently, a Frequent Pattern List (FPL) algorithm, using a simple linear list to store all transactions, was proposed to improve the FP-tree algorithm. However, FPL algorithm still needs to scan database twice. In this paper, an efficient frequent pattern generation algorithm, called FPLI, was proposed to improve the FPL algorithm. FPLI scans the database only once and uses a simple linear list to store all transactions like FPL. By performing simple operations on the list, we can discover the frequent patterns quickly. It is also not necessary for FPLI to rescan database and reconstruct data structure when transaction database is updated or minimum support is varied. Experimental results show that the FPLI algorithm has much better performance than the FPL algorithm.
|
|---|
