Download presentation
Presentation is loading. Please wait.
Published byAlbert McGee Modified over 9 years ago
1
Intelligent Database Systems Lab 國立雲林科技大學 National Yunlin University of Science and Technology SEP/COP: An efficient method to find the best partition in hierarchical clustering based on a new cluster validity index Presenter : Zhen-Feng Weng Authors : Ibai Gurrutxaga, In˜aki Albisua, Olatz Arbelaitz, Jose´ I. Martı´n, Javier Muguerza, Jesu´s M. Pe´ rez, In˜igo Perona 2010/06/30 PR.10 (2010)
2
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 2 Outline Motivation Objective Method Experiments Conclusion Comments
3
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 3 Motivation A cluster hierarchy can offers more information than a single partition. Need post-processing approaches to find a best partition, but most of them are not valid in many cases with non-horizontal cuts.
4
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 4 Objective It proposed a new method called SEP that can searches in an extended partition set. It also proposed a new validity index, COP, to handle the cutting problem in cluster hierarchies. An extended partition set
5
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 5 SEP algorithm An algorithm can find the best partition in the extended partition set of a hierarchy. An Exhaustive search with a threshold ( ≦ 10 or 25 ) A B C D EFGHI 1. C={A} 2. SEP(B),SEP(C) 3.U={B,C} 4. compare(V(C),V(U))
6
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 6 Properties of CVI Partiality: The CVI must be able to evaluate a partial partition. Common CVI: V(P X,X) Partiality CVI: V(P Y,X) Context-independent optimality: The CVI must ensure that if a partial partition is better than another one, it will be better in any context. V(P 1 Y )<V(P 2 Y ) V(P 1 Y ∪ P Z )<V(P 2 Y ∪ P Z ), Y ∩ Z=0
7
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 7 CVIs Calinski–Harabasz (CH): C-Index: Gamma: Davies–Bouldin (DB): Dunn: None of them is with partiality and context- independent optimality.
8
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 8 COP: a new CVI Satisfy the context-independent optimality and partiality properties. COP(root)=COP(leaf)=1 Compute the COP of each node just once
9
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 9 Experiments
10
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 10 Experiments (cont.)
11
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 11 Conclusions It proposed a new post-processing method, called SEP, which can search in a more extensive partition set. And it also define a new CVI, COP for SEP.
12
N.Y.U.S.T. I. M. Intelligent Database Systems Lab 12 Comments Advantage Simple, effective Drawback Application Integrate with SOM
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.