Ant Colony Algorithm for Feature Selection on Microarray Datasets

Development of microarray technology makes the number of research about Bioinformatics will increase as well. Microarray dataset contains genetic information and can be used to analyze thousands of samples and features. Especially in cancer research, the cancer data is generated by microarray technology, and will be a primary data for training and testing in the machine learning process. The main difficulties lie in the nature of microarray gene expression data which usually are noisy and high-dimensional. Microarray dataset usually has a large number of attributes or features, but it has a small number of samples. This condition makes the learning process microarray dataset has become harder because of the curse of dimensionality, where the machine will be difficult to handle a number of data with a very high-dimensional. The solution to handle a high-dimensional dataset and improve the accuracy of microarray dataset is using feature selection. The method of feature selection is followed the principle of natural selection, called Evolutionary Algorithms. We proposed to implement some Evolutionary Algorithms such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Ant Colony Optimization (ACO), which ACO improved F-measure of 9 datasets and ROC Area of 7 datasets from 11 datasets existing in the cancer research.