An Improved Model Based on Negative Selection Algorithm

Abstract

The widely used negative selection algorithm is one of the important algorithms of artificial immune system. However, there are also some disadvantages, such as insufficient learning of self-tolerance in the circumstance of small training set, which affects the detection accuracy. We use a semi-supervised learning mechanism to solve the inadequate learning problem, expand the training sample source, make training to learn more representative samples. Simulation experiments prove that the semi-supervised learning algorithm can improve the training learning process, improve the detection rate, and have strong adaptive capacity.