1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

Title:	Kernel P system-multi objective binary particle swarn optimization feature selection method in microarray cancer dataset
Authors:	Naeimah Elkhani (P72752)
Supervisor:	Ravie Chandren Muniyandi, Assoc. Prof. Dr.
Keywords:	Molecular computers Natural computation Universiti Kebangsaan Malaysia -- Dissertations Dissertations, Academic -- Malaysia
Issue Date:	6-Jul-2018
Description:	According to the statistics, cancer related diseases are the most challenging health problem in all over the world which leads to fatality in the case of late diagnosis. Diagnosis tools based on gene expression profiles have shown significant contribution to the progression of cancer studies. DNA microarrays gene expression is the most commonly used tool because of its ability to monitor few thousand of genes at the same time in their expression level. The main difficulty in this technique is that there are large number of genes (features) compared to the small sample sizes which makes negative impact on the speed and accuracy of technique. Selection of relevant genes is the crucial task for sample classification in microarray data and many research in this field try to extract the smallest group of genes those can provide good diagnosis result. In this study, membrane computing is used to improve accuracy and speed in feature selection and classification methods related to cancer datasets. The proposed model consists of three main part. Firstly, the thesis introduces new criteria to design and develop KP-MObPSO which resembles a wrapper type feature selection. Division rule, rewriting and input/output are used to make an interaction among the genes inside and between the particles. Secondly, an embedded feature selection and classification criteria developed based on KP system. In the second part, the marker gene sets are extracted by the embedded part of the model indicate more stability and reliability based on ROC measure as well as better error rate in compared to wrapper part of the model. Finally, due to the inherent large-scale parallelism feature of membrane computing, any membrane computing inspired model can fully represent this computation model only in the case of using parallel platform. The proposed model applied on the colorectal and breast dataset contains 100 genes with 6 samples. In the first part, KP-MObPSO feature selection model outperforms accuracy of Pure-MObPSO measured by support vector machine (SVM). The proposed KP-MObPSO model implemented on multi-core and graphic processing unit (GPU) to improve the speed of execution. The lowest error rate by embedded model displayed as 0.1111 for breast cancer and 0.0769 for colorectal data. Although the execution of the proposed on multi-core was not able to decrease the time cost significantly, its execution on NVIDA Geforce 680 GPU demonstrates a significant drop of time cost as 164 sec for independent 100 times iterations in compared to 25 min on the central processing unit (CPU) for 25 particles in 100 times iteration. The introduced membrane inspired feature selection and classification method for cancer datasets achieved better performance in terms of accuracy and time cost than pure optimization method.,Ph.D.
Pages:	284
Call Number:	QA76.887.E445 2018 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Information Science and Technology / Fakulti Teknologi dan Sains Maklumat

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