The Improved <i>k</i>-planes Cluster Analysis and Visualization of the Analytical Results
-
- KUROKI Manabu
- The Institute of Statistical Mathematics
-
- YAMASHITA Haruka
- Waseda University
Bibliographic Information
- Other Title
-
- 改良型<i>k</i>-planesクラスター分析法と解析結果の視覚化について
- 改良型k-planesクラスター分析法と解析結果の視覚化について
- カイリョウガタ k-planes クラスター ブンセキホウ ト カイセキ ケッカ ノ シカクカ ニ ツイテ
Search this article
Abstract
<p>This paper proposes an improved k-planes clustering method to classify a p-dimensional data-set into k subsets that are characterized by k different qj dimensional linear structures (0 ≤ qj ≤ p - 1; j = 1, ..., k). The proposed method can be identified with the k-means clustering method when all k subsets can be characterized by the 0 dimensional linear structures, and with the k-planes clustering method when they can be characterized by the same q dimensional linear structures (0 ≤ q ≤ p - 1). In addition, the proposed method is more flexible than those proposed by Bezdek et al. (1981a, 1981b), Bradley and Mangasarian (2000) and Kuroki et al. (2004), in the sense that the same dimensional linear structures for some of k data-sets can be introduced if necessary. Furthermore, letting q∗ = max{qj |j = 1, , ..., k}, we provide a procedure for constructing the p - q∗ dimensional visual hyperplane to visualize the analytical results using the improved k- planes clustering method. Finally, through numerical experiments and a case study, we show that the performance of the improved k- planes clustering method is superior to those of the k-means method and the k-planes method.</p>
Journal
-
- Journal of Japan Industrial Management Association
-
Journal of Japan Industrial Management Association 68 (1), 1-12, 2017
Japan Industrial Management Association
- Tweet
Keywords
Details 詳細情報について
-
- CRID
- 1390282680481570688
-
- NII Article ID
- 130006942289
-
- NII Book ID
- AN10561806
-
- ISSN
- 21879079
- 13422618
-
- NDL BIB ID
- 028159558
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- NDL
- CiNii Articles
- KAKEN
-
- Abstract License Flag
- Disallowed