Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/19660
標題: 一個自我組織映射網路的新視覺化方法
A Novel Visualization Method Based on Self-Organizing Maps
作者: 李亞蓉
Li, Ya-Rung
關鍵字: Self-Organizing Map
自我組織映射圖
Cluster
Visualization
Chi-Square Test
分群
視覺化
神經元
分類
卡方檢定
出版社: 資訊科學與工程學系所
引用: 參考文獻 1. 中文部份 (1) 圖書 [1] 成灝然,統計學,三民書局,民85。 [2] 林惠玲、陳正倉,基礎統計學-觀念與應用,雙葉書廊有限公司,民96。 [3] 徐世輝,應用統計學 (民94),華太文化事業股份有限公司,民94。 [4] 曾憲雄、蔡秀滿、蘇東興、曾秋蓉、王慶堯,資料探勘 Data Mining,旗標出版股份有限公司,民94。 [5] 蘇國樑,統計學,空中大學,民92。 (2) 網路資源 [6] 類神經網路簡介, http://www.im.isu.edu.tw/faculty/pwu/expert/ann.ppt,參考日期2009年2月。 [7] 類神經網路, http://www.im.isu.edu.tw/faculty/pwu/NN/CH06.ppt,參考日期2009年2月。 2. 西文部份 (1) Journal and Conference Articles [8] J. Vesanto and E. Alhoniemi, “Clustering of the self-organizing map,” IEEE Transactions on Neural Networks, Vol.11, No.3, pp.586-600, 2000. [9] R.H. Nielsen, Neurocomputing, Addison Wesley Publishing Company, 1990. [10] R. Helge, M. Thomas, and S. Klaus, Neural Computation and Self-Organizing Maps An Introduction, Addison Wesley Publishing Company, 1992. [11] T. Kohonen, Self-Organizing Maps, 3rd ed., Springer-Verlag Berlin, 1989. [12] T. Kohonen, “The self-organizing feature map,” In Proceedings of the IEEE, Vol. 78, No. 9, pp.1464-1480, 1990. [13] S. Haykin, A Comprehensive Foundation, Prentice-Hall, 1999. [14] A. Ultsch and H.P. Siemon, “Kohonen''s self organizing feature maps for exploratory data analysis,” International Neural Network Conf., pp.305-308, 1990. [15] A. Ultsch, “U*-Matrix: a Tool to visualize Clusters in high dimensional Data,” Philipps-University Marburg, 2004. [16] A. Ultsch, “Maps for the Visualization of high-dimensional Data Spaces,” In Proceedings Workshop on Self organizing Maps, pp. 225 - 230, 2003. [17] J. Vesanto, “SOM-Based data visualization methods,” Intelligent Data Analysis, Vol.3, pp.111-126, 1999. [18] K. Tasdemir, E. Merenyi, and S. Member, “Exploiting Data Topology in Visualization and Clustering of Self-Organizing Maps,” IEEE Transactions on Neuron Networks, Vol.20, No.4, 2009. [19] G. Polzlbauer, M. Dittenbach, and A. Rauber, “Advanced visualization of Self-Organizing Maps with vector fields,” SCIENCE DIRECT on Neuron Networks, Vol.19, 2006. [20] G. Liao, T. Shi, and J. Xuan, A Novel Technique for Data Visualization Based on SOM, Springer-Verlag Berlin, 2005. [21] T. Kohonen, “The self-organizing feature map,” In Proceedings of the IEEE, Vol. 78, No. 9, pp.1464-1480, 1990. [22] A.E. Luang, C.P. Jagdish , K.M. Pramod, and Z. Qin, “A New SOM-based Visualization Technique for DNA Microarray Data,” Conf. on Neural Network, 2006. (2) Electronic Resource [23] Iris Dataset: available on-line. ftp://ftp.ics.uci.edu/pub/machine-learning-database/iris/, Accessed 10 Jan. 2009. [24] Glioblastoma data set: http://en.wikipedia.org/wiki/Glioblastoma_multiforme, Accessed 14 Feb. 2009.
摘要: 自我組織映射網路(Self-Organizing Maps, SOM)在資料分群的運用上是一個很好的方法,而自我組織映射圖視覺化可以讓使用者更容易瞭解資料的特徵。本論文提出一個新的自我組織圖視覺化方法,讓使用者可以輕易的看出資料分群後視覺化的情形,不同群用不同的顏色去做區別,如此一來可以清楚的知道資料分群後分佈的狀況,除此之外,若使用者已知資料標籤,想要瞭解資料分群後每一個神經元中包含資料的類別,本論文的方法也可以清楚呈現資料類別在神經元上的分佈。 在未知資料類別標籤的情形下,想知道資料群聚的分佈狀態,是利用相異群集之邊界資料點較少,或是根本不相鄰的這個特性去找出群與群之間的邊界神經元,進而使視覺化圖可以呈現群聚的位置與邊界。 若是已知資料類別標籤想得知資料在分群後資料種類的分佈狀況,本論文的方法是統計每一個神經元中資料的總個數和每一種類別的資料個數,因此可以得知每一個神經元中資料種類所佔的比例,若神經元中的資料個數大於或等於某一個臨界值,將利用卡方檢定去判別神經元分類的狀況,若小於臨界值則利用機率的大小來判斷神經元該歸屬的種類,進而能夠將每一個神經元做分類,然後再以不同的顏色表示神經元的分類不同。
Self-Organizing Map (SOM) is a very efficient method for data clustering. A visualization tool for SOM can help the user to understand the characteristics of input data. In this thesis, we proposed a novel visualization method by finding decision boundaries among clusters and using different colors for neurons in different clusters. The visualization tool is applicable to data sets with unknown class labels or known class labels. If the class labels are unknown in the data sets as SOM usually handles, two threshold values, namely threshold of cluster boundary and threshold of cluster, are used to determine cluster boundaries and the number of clusters. If the class labels are known in the data sets, the proposed method counts the number of data mapped to a neuron and finds the class distribution in a neuron. If the number of data in a neuron is greater than or equal to a threshold, then Chi-Square test is used to determine the class of the neuron; otherwise, assign the class with highest probability to the neuron. Experimental results show that the proposed method provides better visualization effect for understanding the data distribution in SOM compared to U-matrix and hit histogram.
URI: http://hdl.handle.net/11455/19660
其他識別: U0005-2807200922544800
文章連結: http://www.airitilibrary.com/Publication/alDetailedMesh1?DocID=U0005-2807200922544800
Appears in Collections:資訊科學與工程學系所

文件中的檔案:

取得全文請前往華藝線上圖書館



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.