An Artificial Immune Classification Algorithm based on Particle Swarm Optimization

Abstract

Biological models of the natural immune system have provided the inspiration for artificial immune system, in particular the theories of negative selection, clonal selection and immune networks. The various applications of artificial immune systems have been used for pattern recognition and classification problems, however the traditional artificial immune algorithms have two major problems: the control of the growing of the memory cells population and the major operators of cloning and mutation are taken randomly. In this paper, a new artificial immune classifier based on practical swarm optimization is proposed to solve these problems. The proposed algorithm uses practical swarm to evolve the antibody population. During the training stages, the number of memory cells for each class is predefined to avoid antibody population growing. Then the improved practical swarm optimization method is adopted to evolve antibody population. In each iteration, the antibody as a particle moves to the optimal solution with directional parameters. The new classifier is tested on benchmark datasets using different classification techniques and found to be very competitive when compared to other classifiers.