ANN parametric optimization for cancer classification using best first search and Taguchi method

Md Akizur Rahman (P81293)

Please use this identifier to cite or link to this item: https://ptsldigital.ukm.my/jspui/handle/123456789/476650

Full metadata record

DC Field	Value	Language
dc.contributor.advisor	Ravie Chandren Muniyandi, Assoc. Prof. Dr.
dc.contributor.author	Md Akizur Rahman (P81293)
dc.date.accessioned	2023-10-06T09:23:16Z	-
dc.date.available	2023-10-06T09:23:16Z	-
dc.date.issued	2019-07-03
dc.identifier.other	ukmvital:123559
dc.identifier.uri	https://ptsldigital.ukm.my/jspui/handle/123456789/476650	-
dc.description	Cancer is one of the most significant reasons of death. Immediate detection and diagnosis of cancer is essential for the proper treatment to cure the disease in order to save human lives. Computer-based treatment is very important for cancer classification and proper diagnosis. Today, the artificial neural network (ANN) is being used extensively for diagnosis of cancer classification problems. Sometimes, neural network faces overfitting problem for which the performance of classification is affected. One reason for overfitting is the numerous amounts of neurons in the hidden layer, hence the need for parameter optimization of ANN. Therefore, finding optimal neurons in hidden layer is one of the main reasons for increasing the classification performance in the cancer classification task. For finding the optimal neuron this research utilizes Taguchi method to set the number of the neuron in the hidden layer. By using Taguchi method, this study proposed 15-hidden neuron artificial neural network for cancer classification. Besides, constructing a classifier model for cancer diagnosis and classification where the number of training samples is small relative to the number of features, results in a high dimensional feature space, which may lead to a considerable degradation in the accuracy of the classification due to the “curse of dimensionality phenomenon”. Feature selection (FS) methods play important role in cancer classification. FS methods reduce the dimensionality which can help in increasing the classification ability. Therefore, this study proposes twostep feature selection method for improving neural network classification performance on cancerous dataset. The first step uses Best First Search method for feature extraction while the second step utilizes Taguchi method to find high optimal feature subset from remaining features set. Through the processes, a promising result has been achieved by using breast cancer dataset, colon cancer dataset and ovarian cancer dataset. The result of the experiment revealed the classification accuracy of 99.4% from breast cancer dataset, 99.9% from colon cancer dataset and 99.9% from ovarian cancer dataset. These results are achieved due to the proposed FS method. The achieved performance of this experiment is appeared to be superior in comparison with the other documented research studies regarding cancer classification from the literature on Wisconsin Breast Cancer Dataset, Colon Cancer Dataset and Ovarian Cancer Dataset.,Master of Information Technology
dc.language.iso	eng
dc.publisher	UKM, Bangi
dc.relation	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat
dc.rights	UKM
dc.subject	Universiti Kebangsaan Malaysia -- Dissertations
dc.subject	Dissertations, Academic -- Malaysia
dc.subject	Cancer -- Diagnosis
dc.subject	Cancer classification
dc.subject	Artificial neural network
dc.title	ANN parametric optimization for cancer classification using best first search and Taguchi method
dc.type	theses
dc.format.pages	164
dc.identifier.barcode	005748(2021)(PL2)
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Files in This Item:

File	Description	Size	Format
ukmvital_123559+SOURCE1+SOURCE1.0.PDF Restricted Access		2.54 MB	Adobe PDF	View/Open

Show simple item record Recommend this item