Research Of The Principles And Applications Of Artificial Immune Network Memory Classifier

This paper focuses on the applications of immunity-based computation intelligence in data analysis. On the base of analyzing the faultiness of the present technology based on artificial immune, it presents a new kind of classification technology and compares it with AIRS, which is based on resource limited artificial immune system and also the most successful technology of classification in the field of Artificial Immune System.On the base of describing the principles and mechanisms of natural immune system, the review of the principles and applications of artificial immune system is given in detail. aiNet is a new kind of cluster analysis technology, which is combined with traditional cluster technology. aiNet is improved in viewdata by PCA. It shows that aiNet has many limitations in analyzing high dimensions and multi-classes data and in viewdata. It is not an effective and useful way of data analysis.AIRS is a kind of classifier based on resource limitation artificial immune system and is the most successful classifier at present in the field of Artificial Immune System. It presents a new kind of classifier based on artificial immune network---Artificial Immune Network Memory Classifier. It is tested on the UCI standard data sets and compared with AIRS and the other classical classifiers. The aim is to research the performance of classifier based on artificial immune network. It mainly focuses on the relation between main parameters and the accuracy of classification. The results show that artificial immune network memory can be used to solve the problems of classification. AINMC has good generalized performance and higher accuracy under the conditions of the best set of parameters and adjusting parameters. It has better performance than AIRS and some other traditional classifiers.The analysis to the distributions of memory cells of AIRS and AINMCdiscloses the reason that AINMC has the better performance of classification that of AIRS.The performance of AINMC is tested on multi-dimensions and multi-classes data sets. The performance of it is better than that of AIRS. It is also used to model and predict on non-skewed class distribution, which usually exists in real world. The results are analyzed by ROC and compared with KNN. They show that not only can AINMC resolve the problem of skewed class distribution, but also non-skewed class distribution in real world. Its performance is better than common KNN.AINMC is used in Web text classification in this paper. Based on the test of small examples of Web text, it shows that AINMC can solve well such real world problem. It is also used in information retrieval system. Compared with the other ways of artificial immune system and tradition, it shows that not only can AINMC be used in data mining field to solve data classification in general, but also in Web mining to solve Web text classification.Through the research, it is proved that the classifier based on artificial immune network has better performance than that of the classifier based on resource limitation mechanism. It can be used as a general classifier and also can be used as a new kind of technology of data mining in real world. It introduces new idea of classification to artificial immune system and data mining.