Incremental attribute reduction based on fuzzy dominance neighborhood knowledge granularity in bipolar fuzzy information

Incremental attribute reduction has become a research hotspot in data mining due to its ability to significantly enhance the efficiency of knowledge discovery in dynamic datasets. However, no incremental attribute reduction methods currently exist for bipolar fuzzy decision information systems (BFDIS). Motivated by these issues, this study explores dynamic attribute reduction methods specifically designed for bipolar fuzzy environments. First, a fuzzy dominance neighborhood rough set (FDNRS) model was developed. Next, fuzzy dominance neighborhood knowledge granularity (FDNKG) was defined as an attribute evaluation metric, and a heuristic static reduction algorithm was proposed. Building on this foundation, incremental update mechanisms for FDNKG were derived to handle the dynamic addition and deletion of objects, leading to the development of two incremental attribute reduction algorithms (IFDNKG-AMS and IFDNKG-DMS). Finally, comparative experiments on ten benchmark datasets demonstrated that the proposed algorithms achieved substantial improvements in computational efficiency while maintaining superior reduction quality.