Evolving Complex Network for Classification Problems

In this paper, a new automatic method for constructing and evolving complex network is proposed. This method uses the concept of complex networks (mainly scale-free networks) and combines the essence of immune programming with particle swarm optimization algorithm. The structure of a complex network genotype is evolved using immune programming algorithm with specific parameters, and the fine tuning of the parameters encoded in the structure is accomplished using particle swarm optimization algorithm. The performance of proposed method is compared with flexible neural tree (FNT), neural network (NN), and wavelet neural network (WNN) by using the same breast cancer data set. The results of experimental study indicate that the proposed method is efficient.