Research On Artificial Immune Network Clustering Algorithm Based On Competitive Selection

Mining Data, known as data mining in the academic research, refers to the use of the corresponding algorithm to find useful information from the massive data, which is often used as a part of the database knowledge discovery. Now, clustering is a new method of knowledge discovery. In the previous research, data dimensions are not very high, and the existing clustering methods are able to solve. However, with the rapid advance of science and technology, data is also changing which directly lead to the existing clustering algorithm failure. In this paper, we propose an improved algorithm based on second competition selection, and apply it to the common data sets, UCI data sets and texture image processing. This paper details as follows:(1) On the basis of the theory of aiNet, the concept of second competition selection is put forward. Compared with the original algorithm, the proposed algorithm replaces the network suppression operator and the clonal suppression operator. The new method can improve the anti noise performance and reduce the sensitivity to the noise. In comparison with F-aiNet and aiNet, we conclude that cs-aiNet has a good performance in the processing of noisy data sets. Moreover, compared with the traditional clustering method,the clustering ability of the improved algorithm is not attenuated.(2) In the third chapter, we make the corresponding improvement and add the kernel function method. In the application of complex data sets, the new method often do not get the right results. But after joining the kernel function, the original data sets will become linear, and the processing will becomes simple. This because the nonlinear mapping, data can be transformed from the original space to the feature space. Compared with the other algorithms, the improved algorithm based on kernel function is feasible and effective.(3) A texture image segmentation algorithm based on cs-aiNet is proposed. The proposed method is applied to the segmentation of image processing. In order to reduce the time complexity, we use the gray matrix and wavelet transform to get the gray feature and wave feature of the image, and then we will get the characteristic value as the initial antigen population. Compared with K-means and MOCK, the improved algorithm shows satisfactory segmentation results.