Sparse representation can evaluate the importance of each feature from the overall structure of the data, and mutual information can evaluate the redundancy between features, utilizing the importance and the redundancy propose a supervised feature selection algorithm based on sparse representation and mutual information. Firstly, the algorithm weights 0-1 shows the relationship between the samples; then get low-dimensional embedding space which keeping the structure between sample classification by solving eigenvalue problem; thirdly, using sparse representation to evaluate the contribution degrees of each feature to keep the classification structure of inter-sample, and thus can select features which can maintain preferably the classification structure between the samples; finally, utilizing mutual information to reduce redundancy of the before feature selection results, in order to obtain the final results of the feature selection. In the results of clustering experiment in four different data sets, the method has a better feature selection effect, can achieve better results of features selection at a lower dimension ,meantime, it significantly improves the classification results.
Supervised Feature Selection Based on Sparse Representation and Mutual Information
2023-10-11
2006413 byte
Conference paper
Electronic Resource
English
Semi-supervised sparse feature selection based on multi-view Laplacian regularization
British Library Online Contents | 2015
|Fault feature extraction of rolling element bearings using sparse representation
Online Contents | 2015
|Vehicle Trajectory Analysis System via Mutual Information and Sparse Reconstruction
Transportation Research Record | 2017
|