Simple Method of Increasing the Coverage of Nonself Region for Negative Selection Algorithms

Abstract

One of the intriguing applications of immune-inspired negative selection algorithm is anomaly detection in the datasets. Such a detection is based on the self/nonself discrimination and its characteristic feature is the ability of detecting nonself samples (anomalies) by using only information about the self or regular, samples. Thus the problem space (Universe) is splitted into two disjoint subspaces: One of them contains self samples and the second is covered by the samples which activate the detectors generated by the negative selection algorithms. Hence, the efficiency of negative selection algorithms is proportional to the degree of coverage (by the detectors) of nonself subspace. In this paper, we present a simple method of increasing the coverage for real-valued negative selection algorithm.