| Summary: | 碩士 === 靜宜大學 === 資訊管理學系研究所 === 99 === Periodic pattern mining searches useful periodic patterns in time-related datasets. Previous studies mostly concern the synchronous periodic patterns. Since static transaction database cannot provide the dynamic and timely information to obtain opportune mining results, this study proposes the Dynamic Periodic Pattern Mining model for progressive databases. This study also presents a novel periodic pattern two-dimensional linked list structure to assemble the information of periodic patterns. For each event, the DOEOP algorithm discovers all periodic 1-patterns for all windows of interest (WOI). Finally, the DPPM algorithm can effectively generates all periodic patterns with respect to different WOIs. In this research, we test our proposed scheme and algorithm on two real datasets, Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) and Dow Jones Industrial Average. The experimental results indicate that the proposed algorithm has noticeably good performance.
|
|---|
