Abstract:Artificial immune network clustering is often ineffective when there is noise or undefined cluster boundary in the data. Enlightened by the diversity of immune system, an artificial immune network clustering method based on affinity accumulation was proposed. The method introduced the idea of affinity accumulation and effective evolution strategies into antibodies, and used affinity accumulation in antibodies to describe the distribution trend of spatial density of data. It resulted in a clear cluster structure for the memory network with the effect of secondary immunity. The experimental results show that, the method is effective in clustering while dealing with undefined boundary problems, and is powerful in avoiding noise.