Hierarchical online streaming feature selection based on adaptive ReliefF

Abstract

In hierarchical classification learning, the feature space of data has high dimensionality, and the feature space is unknown with emerging features. To solve the above problems, we proposes an online streaming feature selection algorithm for hierarchical classification based on the adaptive ReliefF. Firstly, the ReliefF is adaptively improved by using the density information of instances around the sample, making it unnecessary to pre-specify parameters. Secondly, the hierarchical relationship between classes is integrated into the algorithm, and a new method for calculating the feature weight of hierarchical data is defined. Then, a feature online importance analysis method is designed based on feature weight. Finally, the adaptive ReliefF algorithm is improved based on feature redundancy, and the feature weight is scaled by the correlation between the features to achieve dynamic feature redundancy update. A large number of experiments verify the effectiveness of the proposed algorithm.