基于實例的模式識別中，由于存在大量實例和特征個數可變特性，導致分類器識別性能低下，難于形成具有高度概括性的共性對象實例。基于免疫計算的概念提取是在有效降低特征個數的同時，提取各類的中心，以此為實例模式對待識別樣本進行分類決策。實驗結果表明算法在保持甚至提高分類精度的同時，不僅有效地降低了特征個數，而且提取的類中心分類效果更好。與基于遺傳算法的概念提取結果相比較，在有限代數內，該算法能收斂到更優的類中心，從而驗證了算法的有效性及其應用潛力。
關鍵詞 免疫計算；概念提取；特征選擇；模式分類

Studies on the Method of Concept Extraction Based on the Immune Computation Liu Jing Zhao Yougang Han Zhongzhi Abstract In the pattern recognition based on examples, the large amounts of examples and the variation of the features cause the low recognition performance of the classifier and make it difficult to draw the succinct generalization of the general examples instance. The concept extraction based on the immune computation is to lower effectively the amount of marks and to extract the center of every class which is used in the classifying of the samples to be classified. The result of the experiment indicates that the algorithm can keep up and even enhance the precision of classification, lower effectively the amount of features and supply a good effect of classifying based on the center of classes. Compared with the result of concept extraction based on the genetic algorithm, this one can converge more effectively to the center of classes in the finite algebra, thus verify the effectiveness and potential application of this algorithm.
Keywords immune computation; concept extraction; feature selection; pattern
classification