Immune Clonal Selection Based Dimension Reduction And Applications

In high-dimension data analysis, no dimension reduction will bring in heavy computation and less probability of valuable information. According to the different ways of obtaining the effective features, dimension reduction can be realized by feature extraction and feature selection. The basic task of feature extraction is to express the sample of high-dimensional space at a lower dimensional subspace by the method of mapping. Feature selection is to select the most effective features with the purpose of reducing the dimension of feature space. Based on researching into the current feature extraction methods, this paper proposes some improved algorithms with intelligent computation methods and techniques. The main contributions can be summarized as follows:1. By researching into classical feature extraction and selection algorithms, the significance of feature extraction and selection is explained. The performance of these algorithms is analyzed and compared through experiments. A new dimension reduction method combining immune clonal selection algorithm and genetic programming is proposed. Immune clonal selection algorithm, which can find the global optimal solution, is used to optimize the polynomial mapping function with tree shape. The experimental results validate the effectiveness of the proposed method.2. A new feature extraction method combining immune clonal selection algorithm and principal component analysis is proposed. In order to overcome the disadvantages of the traditional principal component analysis that the feature vectors provided by PCA are not the most distinguishable, the immune clonal selection algorithm, which can fast converge to the global optimal solution, is applied to select the optimal principal components. Based on the feature vectors selected by the proposed method, the experiment results on hyperspectral remote sensing image show that this method is of effectiveness.3. A new optimal feature extraction method based on immune clonal selection is proposed. Some orthogonal bases are randomly selected as the initial basis vector sets from the original feature space. The direction of the base vectors is optimized to generate the optimal projection vector using the immune clonal selection algorithm. The test results verify the validity of this method. This method provides a new idea of applying the immune clonal algorithm to the optimization problem of feature vector.