A Recursive Regularization Based Feature Selection Framework for Hierarchical Classification
Hong Zhao,Qinghua Hu,Pengfei Zhu,Yu Wang,Ping Wang
DOI: https://doi.org/10.1109/tkde.2019.2960251
IF: 9.235
2021-07-01
IEEE Transactions on Knowledge and Data Engineering
Abstract:The sizes of datasets in terms of the number of samples, features, and classes have dramatically increased in recent years. In particular, there usually exists a hierarchical structure among class labels as hundreds of classes exist in a classification task. We call these tasks hierarchical classification, and hierarchical structures are helpful for dividing a very large task into a collection of relatively small subtasks. Various algorithms have been developed to select informative features for flat classification. However, these algorithms ignore the semantic hyponymy in the directory of hierarchical classes, and select a uniform subset of the features for all classes. In this paper, we propose a new feature selection framework with recursive regularization for hierarchical classification. This framework takes the hierarchical information of the class structure into account. In contrast to flat feature selection, we select different feature subsets for each node in a hierarchical tree structure with recursive regularization. The proposed framework uses parent-child, sibling, and family relationships for hierarchical regularization. By imposing $ell _{2,1}$ℓ2,1-norm regularization to different parts of the hierarchical classes, we can learn a sparse matrix for the feature ranking at each node. Extensive experiments on public datasets demonstrate the effectiveness and efficiency of the proposed algorithms.
computer science, information systems, artificial intelligence,engineering, electrical & electronic