LAUSR

In practical applications, there exist lots of ordered information systems (OISs). In the process of dealing with OISs, dominant preference, which plays a significant role in decision making, should be taken into consideration. With the increasing of data capacity, OISs often evolve with time. In order to extract updated knowledge from evolving ordered data, we have to elaborate computation efforts to re-calculate entire data, which consumes a significant computational cost. Therefore, the computational efficiency is extremely low. In response to this challenge, matrix-based dynamic dominance-based multigranulation rough sets (DMGRSs) approaches, which can improve computational efficiency for updating knowledge, are explored to update multigranulation approximations in dynamic ordered information systems with evolving data. To begin with, we present a matrix representation of dominance-based multigranulation approximations according to the dominant relation matrix and relevant column vectors of each granular structure. Afterwards, the incremental strategies to update dominance-based multigranulation approximations in OISs are proposed when adding or deleting objects. Furthermore, the corresponding dynamic algorithms, which avoid some unnecessary calculations, are explored in DMGRSs. Finally, extensive experiments carried out on nine UCI data sets indicate that the explored dynamic algorithms can achieve promising performance.