Designing artificial neural networks using differential evolution for classifying DNA microarrays

The information obtained from the analysis of DNA microarrays is relevant to identify and predict illness, improve treatments, and to determine which genes are responsible to provoke a specific disease. However, the enormous quantity of genes and the few samples to be analyzed affect the performance of any classifier. For this reason, it is necessary to develop a methodology that combines a robust feature selection technique with a classification algorithm for classifying DNA microarrays. In this paper, we combine a feature selection technique based on the Artificial Bee Colony algorithm with an Artificial Neural Network (ANN). Furthermore, this ANN is automatically designed by a Differential Evolution (DE) algorithm that optimizes the synaptic weights, the architecture, and the transfer functions at the same time. To test the accuracy of the proposed methodology, we use the Leukemia AML-ALL dataset.