Variable Granularity Rough Sets Improve Feature Selection in Multi-Density Data

A variable granularity rough set framework addresses significant uncertainty measurement errors in multi-density data, according to an original research study published in the International Journal of Machine Learning and Cybernetics by Shuang An, Jingyi Kang, and Changzhong Wang. The study establishes the theoretical foundation for Variable Granularity Rough Sets (VGRSs), introducing a novel uncertainty measurement and data reduction framework tailored specifically for datasets with uneven density distributions.

Granular computing theory provides a well-established framework for measuring data uncertainty. In practical applications involving multi-density data, however, disparities in data distribution significantly affect uncertainty measurements. The researchers note that these distribution disparities often cause conventional neighborhood methods to yield substantial errors when measuring data uncertainty.

To resolve this problem, the VGRS model operates from the perspective of variable granularity. For samples exhibiting high classification uncertainty, the framework assigns an initial granule to each sample, detects internal class-label conflicts, and dynamically adjusts the granule. Whether a sample belongs to the lower approximation of a decision class is then determined by its adjusted neighborhood granule.

Building directly upon the VGRS framework, the authors developed feature selection algorithms for semi-supervised data reduction. In comprehensive numerical experiments, the VGRS-based algorithms demonstrated superior performance, which the authors report validates the feasibility, effectiveness, and generalization capability of the proposed model.

Subject of Research: Variable granularity rough sets for semi-supervised feature selection in multi-density data

Article Title: Variable granularity rough sets for semi-supervised feature selection

Article References: An, S., Kang, J., & Wang, C. (2026). Variable granularity rough sets for semi-supervised feature selection. International Journal of Machine Learning and Cybernetics, 17(12), Article 533. https://doi.org/10.1007/s13042-026-03353-5

Image Credits: AI Generated

DOI: 10.1007/s13042-026-03353-5

Keywords: variable granularity rough sets, granular computing theory, semi-supervised feature selection, multi-density data, uncertainty measurement, data reduction, neighborhood methods, classification uncertainty, lower approximation, decision class, class-label conflicts, dynamic granule adjustment, numerical experiments, generalization capability, International Journal of Machine Learning and Cybernetics, Shuang An, Jingyi Kang, Changzhong Wang