cDNA microarray data based classification of cancers using neural networks and genetic algorithms

In this paper, intelligent cancer classification method using cDNA microarray data was developed using neural networks and genetic algorithms. For classification, selection of gene expression data from microarray data are performed as a first step to find highly related genes to disease among microarray data. The fitness values for selection results are determined based on neural network model prediction results. To overcome the limitation of pre-determined number of gene selection method, variable-length chromosome based genetic algorithms are applied. For performance evaluation, pre-tested tumor data are used for classification model and evaluated through blind test. Experimental results are compared with principal component analysis (PCA) based statistical methods, and the test results showed that proposed method has superior classification results (96% accuracy) compare to statistical method (92% accuracy).