Data Visualization by Self-Organizing Map

以SOM做資料視覺化

牧摩度(Momodou Lamim Sanyang)

指導教授 ： 陳朝欽

國立清華大學/電機資訊學院/資訊系統與應用研究所/碩士(2010年)

摘要

Data visualization is very paramount nowadays for the simple fact that we have
acquired huge and complex data and are increasingly accumulating more and more due
to cheap storage devices recently. Most of those data are high-dimensional and
therefore hard for human to visualize. Efforts were made to alleviate this
high-dimensional visualization problem and through researchers endeavor,
Self-Organizing Map (SOM) was born.
The Self-Organizing Map is an unsupervised neural network algorithm that
projects high-dimensional data onto a two-dimensional map which we human can
easily visualize. The projection preserves the topology of the data so that similar data
items will be mapped to nearby locations on the map. It is a powerful method for data
mining and cluster extraction and very useful for processing data of high
dimensionality and complexity.
There are several visualization methods which present different aspects of the
information learned by the SOM to gain insight and guide segmentation of the data. In
this thesis, common visualization methods such as dendrogram, 2d-dendrogram,
principal component projection, label of maps, U-matrix and some recently introduced
methods such as P-matrix and the U*-Matrix plots are used to visualize the results on
four data sets: IRIS which has 150 patterns with 3 classes, each class has 50 patterns,
each pattern has four features; 8OX has 45 patterns with 3 classes, each class has 15
patterns, each pattern has 8 features; A microarray data set ALL-AML Leukemia with
38 patients of 2 classes (27 ALL, 11 AML), each patient has 7129 genes; and Colon
Tumor with 62 samples (22 normal, 40 tumor) of 2 classes with a total of 2000 genes.
The visualization results of each of these data sets are reported using the
aforementioned methods, the 2d-dendrogram method seems to be a better tool for
visualizing the microarray data and all the methods perform well on the IRIS and 8OX.

關鍵字

SOM ； 資料視覺化

並列摘要

無資料

並列關鍵字

SOM ； Data Visualization

參考文獻

Proceedings of National Academy of Sciences of the United States of America, vol.

96, 6745-6750, 1999.

Blackwell, Cluster Analysis in Marketing Research, 1994.

[Bald2002] P. Baldi and G. Hatfield, “DNA microarrays and gene expression”, Cambridge

relevance in feature selection for microarray data”, IEEE Intelligent