Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps

碩士 === 國立雲林科技大學 === 資訊管理系 === 103 === The self-organizing map (SOM) is a useful tool which has visualization capability in the exploratory phase of data analysis. It can explore deeper relationship between data by clustering on two-dimensional space. Many SOM clustering methods have been proposed, s...

Full description

Bibliographic Details
Main Authors:	Yan-Shou Sie, 謝彥守
Other Authors:	Chung-Chian Hsu
Format:	Others
Language:	en_US
Published:	2015
Online Access:	http://ndltd.ncl.edu.tw/handle/12887931620875808235

id	ndltd-TW-103YUNT0396040
record_format	oai_dc
spelling	ndltd-TW-103YUNT03960402016-07-02T04:21:20Z http://ndltd.ncl.edu.tw/handle/12887931620875808235 Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps 應用以拓撲基礎的階層式分群方法改善自我組織映射圖的分群效能 Yan-Shou Sie 謝彥守碩士國立雲林科技大學資訊管理系 103 The self-organizing map (SOM) is a useful tool which has visualization capability in the exploratory phase of data analysis. It can explore deeper relationship between data by clustering on two-dimensional space. Many SOM clustering methods have been proposed, such as hierarchical, partitioning, density-based and connectivity strength. In particular, the connectivity strength clustering algorithm (CONN) was proposed recently and achieved good performance on SOM clustering. CONN is a data-topology-based hierarchical agglomerative clustering method which can automatically determine the number of clusters by using cluster validity indices. However, CONN still has some problems due to its characteristics. As a result,the CONN clustering can’t take the best number of clusters. In this study, we propose a new approach by combining the Euclidean distance with CONN in similarity measure to solve the problems. Experimental studies on synthetic datasets showed that our hybrid similarity measure yields higher accuracy on SOM clustering. Chung-Chian Hsu 許中川 2015 學位論文 ; thesis 33 en_US
collection	NDLTD
language	en_US
format	Others
sources	NDLTD
description	碩士 === 國立雲林科技大學 === 資訊管理系 === 103 === The self-organizing map (SOM) is a useful tool which has visualization capability in the exploratory phase of data analysis. It can explore deeper relationship between data by clustering on two-dimensional space. Many SOM clustering methods have been proposed, such as hierarchical, partitioning, density-based and connectivity strength. In particular, the connectivity strength clustering algorithm (CONN) was proposed recently and achieved good performance on SOM clustering. CONN is a data-topology-based hierarchical agglomerative clustering method which can automatically determine the number of clusters by using cluster validity indices. However, CONN still has some problems due to its characteristics. As a result,the CONN clustering can’t take the best number of clusters. In this study, we propose a new approach by combining the Euclidean distance with CONN in similarity measure to solve the problems. Experimental studies on synthetic datasets showed that our hybrid similarity measure yields higher accuracy on SOM clustering.
author2	Chung-Chian Hsu
author_facet	Chung-Chian Hsu Yan-Shou Sie 謝彥守
author	Yan-Shou Sie 謝彥守
spellingShingle	Yan-Shou Sie 謝彥守 Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
author_sort	Yan-Shou Sie
title	Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
title_short	Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
title_full	Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
title_fullStr	Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
title_full_unstemmed	Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps
title_sort	apply topology-based hierarchical clustering to improve clustering performance of self-organizing maps
publishDate	2015
url	http://ndltd.ncl.edu.tw/handle/12887931620875808235
work_keys_str_mv	AT yanshousie applytopologybasedhierarchicalclusteringtoimproveclusteringperformanceofselforganizingmaps AT xièyànshǒu applytopologybasedhierarchicalclusteringtoimproveclusteringperformanceofselforganizingmaps AT yanshousie yīngyòngyǐtàpūjīchǔdejiēcéngshìfēnqúnfāngfǎgǎishànzìwǒzǔzhīyìngshètúdefēnqúnxiàonéng AT xièyànshǒu yīngyòngyǐtàpūjīchǔdejiēcéngshìfēnqúnfāngfǎgǎishànzìwǒzǔzhīyìngshètúdefēnqúnxiàonéng
_version_	1718333536986988544

Apply Topology-Based Hierarchical Clustering to Improve Clustering Performance of Self-Organizing Maps

Similar Items