Abstract

The dynamic tendency of network environment determines that system can achieve an accurate fault diagnosis only by self-learning. Inspired by characters of artificial immune and adaptability of dynamic clonal selection algorithm for dynamic environment, an immune algorithm applied to network fault diagnosis was proposed based on the detector population quality and the memory characteristics. The clonal expansion strategy was designed to improve the quality of mature detector populations and the classification memory strategy can achieve dynamic updated memory detector population through evaluating the effectiveness of the memory detectors. The experimental results show that the network fault diagnosis based on immune theory can achieve successive learning to accommodate the emerging new situations, and improve the accuracy rate and efficiency of detecting known and unknown faults.