Abstract:

To resolve the poor performance of classifiers owing to the irrelevant and redundancy features, a feature selection  algorithm based on approximate Markov blanket and dynamic mutual information is proposed, then it is introduced to an ensemble feature selection algorithm. In the ensemble algorithm, a base classifier is trained based on Bagging and the  proposed feature selection algorithm, and the base classifier diversity is introduced to selective ensemble. Finally, the  weighted voting method is utilized to fuse the base classifiers’ recognition results. To attest the validity,  experiments on data sets with support vector machine (SVM) as the classifier are carried out. The results have been  compared with singleSVM, BaggingSVM and ABSVM. Experimental results suggest that the proposed algorithm can get  higher classification accuracy.