Improved Artificial Immune Network And Its Application In Data Processing

With the development of information technology, more and more people solve and analyze the problems from the perspective of artificial intelligence.Artificial immune system formed a set of unique theory by simulating natural immune mechanism, with the self-organizing, learning, memory, distributed, etc.In recent years,with the continuous deepening of the scholars on the research of artificial immune system,artificial immune system gradually shows its superiority in the field of data mining.The artificial immune system is introduced into the field of data mining has become a hotspot in recent years.In this paper,base on the theory of the artificial immune,the research content is divided into the following two parts:(1)The inadequacy of existing artificial immune network algorithm(aiNet) was improved and proposed a dynamic adaptive immune network base on the niche technology and fuzzy extraction for prior knowledge(NFDA-aiNet). Base on aiNet,The algorithm can dynamically control its parameters, including the selection of antibodies to clone and the immune suppression threshold among the antibodies,so that they can adapt to the dynamic changes of the immune network. In addition the traditional affinity function of evolution is mainly based on the distance, which can not evaluate the affinity globally. The niche technology was introduced in this paper,on that the affinity function was defined based on niche technology instead of the simple distance function. unliked the traditional immune network,the proposed algorithm first call the Fuzzy C-Means clustering on the original data to generate the initial antibody set,which can effectively promote the learning efficiency of the immune evolution.The experiments use 4 data sets of UCI to conduct a comprehensive test on the classification accuracy,the clustering performance and the network compression ability of the NFDA-aiNet algorithm.The experimental results show that NFDA-aiNet has a higher classification accuracy,and the network convergence speed is greatly increased.In follow-up studies,several key parameters of NFDA-aiNet were test and two random convergence were used to analyze the convergence of NFDA-aiNet.(2)On the basis of the traditional artificial immune network, typical strategies of the multi-agent technology are integrated into the evolution of the immune network. The neighborhood clone selection was introduced into the algorithm, make the process of operation from local to the whole, which can simulate the natural evolution model of the immune network more comprehensive. In addition the neighborhood competition and neighborhood collaborative operators between the antibody were introduced into the evolution process of the immune network, which can improve the dynamic analysis capabiility of the network.Classification experiments use the UCI data and a remote sensing image of Zhangzhou sea area, experiment obtain the overall classification accuracy of 92% and Kappa coefficient of 0.91, UCI data also get a good result. Results show that the algorithm is an effective classification algorithm.