Behaviour of classification algorithms for cancer detection based on different attribute sets using microarray gene expression data

  • Simardeep Kaur, Dr. Maninder Singh
Keywords: Gene expression data, Cancer classification, Leukaemia etc.

Abstract

Microarray technology is bestowed with the strengths to offer gene expression data in terms of expression profile of thousands of genes across biological processes. The most exciting artefact of the microarray assay reflects the distinction of cancers based on the patterns exhibited by the gene expression data. In the present work, knowledge gained through microarray assay is used for the classification of leukaemia into Acute Myeloid Leukaemia (AML) and Acute Lymphoblastic Leukaemia (ALL). In the present work, the challenges of feature extraction is addressed with Principal Component Analysis (PCA), Canonical Correlation (CC) and Cosine Correlation (CosC) and their effectiveness  is evaluated with the involvement of three computational intelligence approaches namely, Support Vector Machine (SVM), Linear Discriminant Analysis (LDA) and Feed Forward Neural Network (FFNN) separately at classification stage. The performance of the implemented combinations is determined in terms of True Positive Rate (TPR), False Positive Rate (FPR), Kappa-Coefficient (KC) and accuracy. The simulation analysis of all the nine combinations is performed against 500 samples representing gene expression data for leukaemia. Experimentation has shown that PCA with FFNN outperformed all the other combinations in terms of TPR, FPR, KC and accuracy with an average classification accuracy of 0.6231.

Published
2021-07-23
How to Cite
Dr. Maninder Singh, S. K. (2021). Behaviour of classification algorithms for cancer detection based on different attribute sets using microarray gene expression data. Design Engineering, 4551- 4577. Retrieved from http://thedesignengineering.com/index.php/DE/article/view/2900
Section
Articles