Classification Algorithm Based On Artificial Immune Network

Along with the development of information technology, the method of computational intelligence in the application of data mining has become a focus of the study. Artificial immune system has such features as learning, memory, self-organization, distributed etc., by simulating mechanisms of the natural immune system. If the artificial immune algorithm, especially the artificial immune network algorithm is properly applied to the classification learning and recognition algorithms, the learning and memory capacity of classification algorithm can be enhanced to improve their recognition accuracy. Based on the in-depth analysis of the immune mechanism, two kinds of classification algorithm based on artificial immune network are proposed in this paper.First of all, the paper introduces the immune system and classification in the data mining briefly, including the theory of natural immune system, artificial immune system mechanism, machine learning and the existing classification algorithm in more details. Then, the paper presents a artificial immune classification algorithm with negative selection mechanism and a artificial immune classification algorithm based on adaptive radius respectively, experimental analysis are conducted respectively, too.The classification algorithm with negative selection mechanism based on artificial immune network introduces the mature mechanism of B cells in the thymus to the artificial immune network, so that the co-evolution of the antibodies of different types makes the final network of antibodies can represent the typical samples, reducing the number of memory cells and obtaining better classification result.The classification algorithm based on adaptive radius and artificial immune network introduces the multi-granularity into algorithm, the algorithm retains the density information of training data through adjusting the inhibiting radius of antibodies adaptively and introduces a mechanism of radius attenuation and feedback, the datas failed to be learned will be relearned in a smaller granularity, in order to obtain the more representative memory cells.