Research of the classification feature selection based on artificial immune computation

Feature selection is the important research in the current field of information, particularly the field of pattern recognition. In order to reduce the dimension of attributes and improve the classification accuracy, a algorithm of the classification feature selection based on artificial immune computation (CFSAIC) was proposed. Did simulation experiments with the data sets of Ionosphere, wdbc and pima-indians-diabetes respectively, selected a group of optimum feature subset, then constructed classifier with KNN Rule. It shows that the algorithm of feature selection based on artificial immune computation reduces the data dimension, after constructing classifier with KNN, it improves the classification accuracy comparing to KNN, and also higher than the algorithms of AINC, CSA, MVIN and FCM.