binary particle swarm optimization

Microarray data have an important role in identification and classification of the cancer tissues. In cancer researches always a few samples of microarrays are led to some problems in designing the classifiers، so non-informative genes have been removed from microarray data before classification using the preprocessing gene selection techniques. Basically، appropriate gene selection method can significantly improve the performance of cancer classification. In this paper، a new method is proposed based on hybrid model Binary Particle Swarm Optimization algorithm and Bayesian Linear Discriminant Analysis in order to classification of large scale microarray data. First، the position of each particle is represented in the form of binary vector and random، as each bit illustrates a gene. The zero and one bits represent that the corresponding feature (gene) is not/is selected، respectively. So the position of each particle clarifies a gene subset and fitness of each particle is calculated using Bayesian Linear Discriminant Analysis algorithm to quality evaluation of selected gene subset by that particle. The proposed algorithm is applied on four cancer datasets and its results are compared with other existed methods. Simulation results illustrate that proposed algorithm has high accuracy and validity compared to other existed methods and enables to select the small subset of informative genes in order to increase the classification accuracy.