|
Title:
|
CLASSIFICATION OF CANCER MICROARRAY DATA USING NEURAL NETWORK |
|
Author(s):
|
Myungsook Klassen |
|
ISBN:
|
978-972-8924-30-0 |
|
Editors:
|
Nuno Guimarães and Pedro Isaías |
|
Year:
|
2007 |
|
Edition:
|
Single |
|
Keywords:
|
micro array data, attribute reduction, back propagation neural networks, shrunken centroid approximation, Levenberg-Marquardt Algorithm |
|
Type:
|
Full Paper |
|
First Page:
|
127 |
|
Last Page:
|
134 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Classification of four different types of the SRBCT cancer was studied in this paper. The typical micro array data has a small number of samples and a very large number of genes. The SRBCT cancer data has over 2300 genes and 88 samples. For efficient classification with a neural network, the size of genes was reduced. The reduction method uses concept of the nearest shrunken centroid, but in a simpler fashion to avoid the iterative process of threshold computation of the shrunken centroid. With the proposed reduction method, 52 genes were selected, among which forty seven genes responded to only one cancer type. For classification, the back propagation neural network with Levenberg-Marquardt (LM) algorithm was used for efficient learning. It was hoped that the important characteristic of neural networks such as generalization of input samples, was compensating for inputs genes selected by the proposed simple approximation method. The experimental results of 100% classification rates supported the claim. Comparison of results using back propagation neural network with and without the LM algorithm are also presented. |
|
|
|
|
|
|
