Classification Algorithms for Cancer Detection Using Microarray Gene Expression Data and Multiple Attribute Sets

Abstract

The ability to provide gene expression data in the form of expression profiles of thousands of genes across biological processes is one of the advantages of microarray technology.  Based on the patterns shown by the gene expression data, the microarray assay's most intriguing artifact represents the differentiation of tumors.  The current study uses the knowledge gathered from the microarray assay to classify leukemia into two categories: acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML).  Three computational intelligence approaches—Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Feed Forward Neural Network (FFNN)—are used in this work to assess the efficacy of Principal Component Analysis (PCA), Canonical Correlation (CC), and Cosine Correlation (CosC) in addressing the difficulties of feature extraction. Accuracy, True Positive Rate (TPR), False Positive Rate (FPR), and Kappa-Coefficient (KC) are used to evaluate the performance of the combinations that have been put into practice.  All nine combinations are subjected to a simulated examination using 500 samples that reflect leukemia gene expression data.  With an average classification accuracy of 0.6231, experiments have demonstrated that PCA with FFNN performs better than all other combinations in terms of TPR, FPR, KC, and accuracy.