Publications-NSC Projects
Article View/Open
Publication Export
-
Google ScholarTM
NCCU Library
Citation Infomation
Related Publications in TAIR
題名 多分類器系統建構之理論與實務 作者 徐國偉
Hsu,Kuo-Wei貢獻者 國立政治大學資訊科學系 關鍵詞 多分類器系統;分類器差異性;分類器整合
Multiple Classifier System;Multi-Classifier System;Ensemble;Diversity;Classifier Combination日期 2012 上傳時間 11-Oct-2013 15:13:07 (UTC+8) 摘要 本計畫旨在發展多分類器系統建構之理論與實務。近年來在 資料探勘領域，多分類器系統是一個越來越有吸引力的研究 主題且被應用在各式各樣的領域。然而，我們對其基礎仍然 缺乏全面了解。我們可能知道哪個多分類器系統是有效的， 但是我們卻不清楚它為什麼有效，而且我們對於如何有效地 建構一個有效的多分類器系統的知識是有限的。雖然有許多 成功的案例，但是在更多案例裡面多分類器系統不被認為是 成功的。因此，能夠解釋為何一個多分類器系統是不成功的 與能夠解釋為何一個多分類器系統是成功的是同等重要。再 者，大多數多分類器系統採取的是針對特定問題或資料的特 別設計，然而這樣的設計嚴格限制了他們可能的延伸與應 用。如何設計一個泛用型的多分類器系統是一個重要的研究 課題。在本計畫中，我們將採用理論與實證研究去對多分類 器系統的建構做深入的探討，並聚焦於以異質分類演算法建 置的多分類器系統。本計畫可對國內外的多分類器系統研究 與應用做出實質貢獻，而部分研究成果已發表於專業國際會 議。
The purpose of this project is to develop theory and practice of the construction of Multiple Classifier System. Recently, in the data mining field, Multiple Classifier System has been an increasingly attractive research topic and been applied to various domains. However, we still lack a comprehensive understanding of its foundation. We may know which Multiple Classifier System is effective but we are not clear why it is effective, and we have limited knowledge of how to effectively construct an effective Multiple System. Although there are many success cases, there are even more cases where Multiple Classifier System is not considered success. Therefore, being able to explain why a Multiple Classifier System is unsuccessful is equally important to being able to explain why a Multiple Classifier System is successful. Furthermore, most systems are ad-hoc designs, i.e. they are designed specifically to some problems or data sets, while such designs severely restrict their possible extensions and applications. For that reason, a critical research topic is how to design a general-purpose Multiple Classifier System. In this project, we plan to use theoretical and empirical study to have an in-depth of the construction of Multiple Classifier System, and we will focus on the Multiple Classifier Systems built on heterogeneous classification algorithms. This project could make practical contributions to research and applications of Multiple Classifier System, and part of the results have been published on professional international conferences.關聯 執行期間：10008-10107 資料類型 report dc.contributor 國立政治大學資訊科學系 en_US dc.creator (作者) 徐國偉 zh_TW dc.creator (作者) Hsu,Kuo-Wei - dc.date (日期) 2012 en_US dc.date.accessioned 11-Oct-2013 15:13:07 (UTC+8) - dc.date.available 11-Oct-2013 15:13:07 (UTC+8) - dc.date.issued (上傳時間) 11-Oct-2013 15:13:07 (UTC+8) - dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/61318 - dc.description.abstract (摘要) 本計畫旨在發展多分類器系統建構之理論與實務。近年來在 資料探勘領域，多分類器系統是一個越來越有吸引力的研究 主題且被應用在各式各樣的領域。然而，我們對其基礎仍然 缺乏全面了解。我們可能知道哪個多分類器系統是有效的， 但是我們卻不清楚它為什麼有效，而且我們對於如何有效地 建構一個有效的多分類器系統的知識是有限的。雖然有許多 成功的案例，但是在更多案例裡面多分類器系統不被認為是 成功的。因此，能夠解釋為何一個多分類器系統是不成功的 與能夠解釋為何一個多分類器系統是成功的是同等重要。再 者，大多數多分類器系統採取的是針對特定問題或資料的特 別設計，然而這樣的設計嚴格限制了他們可能的延伸與應 用。如何設計一個泛用型的多分類器系統是一個重要的研究 課題。在本計畫中，我們將採用理論與實證研究去對多分類 器系統的建構做深入的探討，並聚焦於以異質分類演算法建 置的多分類器系統。本計畫可對國內外的多分類器系統研究 與應用做出實質貢獻，而部分研究成果已發表於專業國際會 議。 - dc.description.abstract (摘要) The purpose of this project is to develop theory and practice of the construction of Multiple Classifier System. Recently, in the data mining field, Multiple Classifier System has been an increasingly attractive research topic and been applied to various domains. However, we still lack a comprehensive understanding of its foundation. We may know which Multiple Classifier System is effective but we are not clear why it is effective, and we have limited knowledge of how to effectively construct an effective Multiple System. Although there are many success cases, there are even more cases where Multiple Classifier System is not considered success. Therefore, being able to explain why a Multiple Classifier System is unsuccessful is equally important to being able to explain why a Multiple Classifier System is successful. Furthermore, most systems are ad-hoc designs, i.e. they are designed specifically to some problems or data sets, while such designs severely restrict their possible extensions and applications. For that reason, a critical research topic is how to design a general-purpose Multiple Classifier System. In this project, we plan to use theoretical and empirical study to have an in-depth of the construction of Multiple Classifier System, and we will focus on the Multiple Classifier Systems built on heterogeneous classification algorithms. This project could make practical contributions to research and applications of Multiple Classifier System, and part of the results have been published on professional international conferences. - dc.format.extent 2381107 bytes - dc.format.mimetype application/pdf - dc.language.iso en_US - dc.relation (關聯) 執行期間：10008-10107 en_US dc.subject (關鍵詞) 多分類器系統;分類器差異性;分類器整合 en_US dc.subject (關鍵詞) Multiple Classifier System;Multi-Classifier System;Ensemble;Diversity;Classifier Combination en_US dc.title (題名) 多分類器系統建構之理論與實務 zh_TW dc.type (資料類型) report en
