OFHR: Online Streaming Feature Selection With Hierarchical Structure Based on Relief

Online feature selection for hierarchical classification learning based on improved ReliefF

“…To demonstrate the effectiveness of the OFHIR algorithm, this section selects five online streaming feature selection algorithms: OFS‐A3M, 25 Fast‐OSFS, 39 OSFS, 39 SAOLA, 40 and OFS‐Density 41 . Meanwhile, two online streaming feature selection algorithms for hierarchical classification learning: KFOHS 27 and HSFSAR 29 are selected. Among them, the significance levels of Fast‐OSFS (abbreviated as FOSFS), OSFS, and SAOLA are all set to 0.01, OFS‐Density (abbreviated to OFSD) sets $$$ \alpha $$$ to 0.05, HSFSAR sets $$$ \lambda $$$ = 0.0001 and the parameter $$$ \delta $$$ in KFOHFS is set to 0.0001, respectively.…”

“…It can be seen from Tables 3–6 that the average performance of the OFHIR algorithm all rank first under the four evaluation indicators of Accuracy, H‐F1, LCA‐F1, and TIE. Compared with HSFSAR, 29 OFHIR takes into account the interactivity between features. Therefore, the performance of the algorithm will be improved or flat on most datasets, especially when the number of samples is too large, such as in the Cifar and VOC datasets, the performance under the four evaluation indicators was significantly improved.…”

“…Transforming the feature selection problem under static and fixed conditions into an online streaming feature selection problem under the concept of flow features plays a key role in dealing with such practical problems. At present, many scholars have conducted in‐depth research on the flow feature selection 25‐29 . Zhou et al 25 defined an adaptive density neighborhood relation without prior information based on neighborhood rough set, and proposed a new online streaming feature selection algorithm.…”

“…Wang et al 28 applied the hierarchical relationship between nodes in a tree structure to the Relief algorithm, but this algorithm required to specify the relevant parameters. Wang et al 29 proposed an online streaming feature selection algorithm for hierarchical classification based on the adaptive ReliefF algorithm, this algorithm did not need to specify parameters in advance, but did not consider the interaction between features. In addition, most of the existing hierarchical classification algorithms do not consider the dynamics of features, while traditional online streaming feature selection algorithms often ignore the hierarchical structure between classes.…”

“…Finally, the correlation between the features is used to scale the feature weight, so that the adaptive ReliefF algorithm can update the redundant features, and a new online streaming feature selection algorithm is constructed for hierarchical classification. Different from HSFSAR, 29 the proposed algorithm takes into account the interactivity between features, and a new online correlation analysis method is proposed, which considers not only the correlation between features and classes, but also the interaction between features. A new online hierarchical feature selection framework is constructed.…”