Research on Principal Components Weighted Based on Real-valued Negative Selection Algorithm

Abstract

In order to improve the identification and distribution performance of the detector, this paper proposes Principal Component Weighted Real-valued Negative Selection Algorithm(PCW-RNS) which is based on principal component weighting. The similarity between this algorithm and the classical real-valued detector generating algorithm based on generation-and-elimination lies in the fact that neither adopt any optimization method to optimize the performance of the detector, but only relying on the detection performance of the detector to detect anomalies. Because of the irrelevance between the principal components and the application of weighted Euclidean distance as the matching rules, the detector can adjust its radius according to the distribution of non-self space, thus obtaining higher detection rate of the detector and improving distribution performance of the detector. In this way, we can not only better the identification performance of the detector and obtain a higher detection rate, but also effectively reduce the false alarm rate.