An Improved Artificial Immune Network Based On The Secondary Immune Mechanism For Data Clustering
Yangyang Li,Xiaoju Hou,Licheng Jiao,Yu Xue
DOI: https://doi.org/10.1007/978-3-319-68505-2_45
2017-01-01
Abstract:Data clustering is a typical method in data mining. As a effective algorithm for clustering, the Artificial Immune Network is inspired by natural immune system can reflect the structure of the given dataset, filter redundancy and cluster datasets without the number of clusters, so far it is widely used. However, it can't effectively identify the noise nodes, the running time is long and too much parameters are set in improved algorithms. In order to shorten running time and reduce the impact of parameters, this paper proposes an improved artificial immune network based on the secondary immune mechanism. The Clone operator and Mutation operator are replaced by Competition Selection operator and Competition Selection strategy, which are inspired by the resource limited artificial immune system. Because the algorithm can reach a stable convergence only through two times, so it greatly reduce the running time; and can effectively identify the noise nodes due to the introduction of stimulation level. A number of datasets including artificial datasets and real-world datasets are used to evaluate the performance of the proposed algorithm and the other existing clustering algorithms, such as K-means, FCM, SC, aiNet and FCAIN. The simulation results indicate that the proposed artificial immune network algorithm is an effective and efficient method in data clustering.