K5. Merging genetic algorithm with different classifiers for cancer classification using microarrays

Cancer classification is considered a big challenge for many researchers due to its role in the management of the disease. Thanks to the microarray technology, this task can be automated instead of laboratories procedure. To guarantee accurate classification, microarray datasets must be reduced using gene selection techniques to extract only the informative genes. Genetic algorithm offers a hope to select subsets of these genes from the original microarray datasets. In this paper, genetic algorithm, as a pre-processing step, is merged with three different classifiers forming three classification systems. Two of these classifiers are proposed to be used as the genetic algorithm fitness function. Also, two different methods are proposed for choosing the genetic algorithm initial population. The results of the thee systems were promising as two of them reached perfect classification on one dataset using only four genes and very reliable classification accuracy on the other two datasets.