Feature Selection Using Generalized Multi-Granulation Dominance Neighborhood Rough Set Based on Weight Partition

Rough set theory, as an academic hotspot in the field of artificial intelligence, has provided a solid theoretical foundation for feature selection. However, with the continuous updating of large datasets, classical rough set theory is no longer applicable. Multi-granulation rough set theory is an extension of rough set theory that can better handle complex datasets. Therefore, this paper proposes a generalized multi-granulation dominance neighborhood rough set model based on weight distribution and discusses some relevant properties of this model. Furthermore, a new information entropy is constructed based on this model to handle uncertainty in data. This approach enhances the ability to describe uncertainty and enables more effective feature selection. As a result, a forward heuristic feature selection algorithm is developed to find the optimal feature subset. Finally, the effectiveness of the proposed algorithm is validated through instance analysis on twelve publicly available datasets.