Research Of Feature Selection Algorithm And Its Application Based On Immune Computation

In recent years, with the rapid development of computer technology and networktechnology, data and information stored in electronic form increases quickly, and thedata was no doubt valuable resource. However, with the amount of informationcontinues to expand and the requirements and recognition accuracy gradually improved,the techniques of optimizing feature selection are particularly important. An efficientfeature selection algorithm can not only provide strong protection for the classification、decision-making, but also reduce the overhead required for the task. To analyze the vastamounts of crude data, dig out useful knowledge hidden in the data, data miningtechniques were needed. Inspired by biological immune system that with strong featureextraction、learning、rapid evolution and memory characteristics and so on, whichpursued in data mining。First, the background of a feature selection algorithm based on immunecomputation was introduced, and it made a summary description of the artificialimmune system and its application, and described some feature selection algorithms.For information resource model, feature selection algorithm and dimension reduction,the key technologies relating to classification were described, in this launched theresearch of feature selection algorithm and its application based on immunecomputation.After simply introduced the biological principles of artificial immune system-themechanism of biological immune system, artificial immune system algorithm wereintroduced, focusing on the artificial immune system to learn from the natural immunesystem with the characteristics and has been successfully applied in many areas. Featureselection algorithm based on immune computation was put forward in the paper, theexperiment for feature selection on standard data set was done and then combined withk nearest neighboring (KNN) algorithm to structure classifier, and then compared withKNN classifier and immune classifier proposed by other scholars, the purpose was tostudy the performance of feature selection algorithm based on immune computation,and then applied in the assisted medical diagnostic system. It focused on the algorithmof feature selection based on immune computation and how to apply it in auxiliarymedical diagnosis and its effectiveness. The experimental results showed that artificialimmune algorithm could apply in the feature selection of classification. Feature selection based on immune computation didn’t check up each feature combination, toavoid pre-determining the number of characteristics, it selected a set of valid featurecombination with some intelligent random search strategy according to the affinityevaluation function. In the case of suitable parameters, strong feature extraction wasreflected by testing on standard data sets, and dimensionality reduction as well asbiological immune system’s characteristics like self-learning. The classified accuracy ofthe algorithm combined with KNN rules that tested on standard data set was higher thanFuzzy C-Means algorithm, multi-valued immune algorithm, clone selection based onclassification and artificial immune network classification algorithm. For the assistedmedical diagnostic system, the accuracy of disease classification could reach97%, andprovided a powerful analytical tool for expert to diagnose patients.