An OWA-based multi-granulation fuzzy rough set model using Choquet integrals and its applications

Rough set theory has been widely utilized in the field of intelligent information processing for representing and managing uncertain knowledge. However, most existing fuzzy rough set models rely on simple logical operators (e.g., ⋁ and ⋀) to aggregate information from multiple granules, which limits their flexibility and adaptability in complex data environments. This paper proposes a new multi-granulation fuzzy rough set model based on ordered weighted averaging (OWA) approaches by using Choquet integrals. The model leverages the flexibility of OWA aggregation and the nonlinear integration capability of Choquet integrals to effectively fuse information from multiple fuzzy relations. First, a new multi-granulation fuzzy rough set model is established using Choquet integrals with fuzzy measures. When the fuzzy measure is symmetric, the model simplifies to an OWA-based aggregation of fuzzy rough approximations, providing a more flexible and powerful framework for multi-granulation fusion. Next, the relevant characteristics of the model are explored. The presented model satisfies almost all the properties of the classical rough set model. Several innovative concepts for data mining are introduced, including fuzzy rough integral positive region, fuzzy rough integral dependency, and integral dependency reduction. Finally, a quick feature selection algorithm under this new model is proposed. Additionally, the proposed approach is evaluated against existing methods in classification tasks, demonstrating its feasibility and effectiveness through several public datasets.