Dynamic updating approximations approach to multi-granulation interval-valued hesitant fuzzy information systems with time-evolving attributes

Since data is furious growth and rapid alteration, it is completely imperative to monitor and update real-time data promptly. There is no denying that calculating approximations by means of classical approach is pretty time-consuming for an information system with attribute sets varying constantly. Whereas, dynamic updating approximations method takes full advantage of previous knowledge instead of calculating from scratch, which saves a large amount of time. Enlightened by this idea, our work focuses on researching mechanisms of dynamic updating approximations caused by the variation of attributes in multi-granulation interval-valued hesitant fuzzy information system (MG-IVHFIS). To begin with, the average dominance relation which reduces the restriction of universal dominance relation in reality is recommended, then an average dominance rough set based on this relation is established in MG-IVHFIS. Additionally, we study four mechanisms for updating approximations from the perspective of optimism and pessimism in dynamic MG-IVHFIS when some attributes are removed or inserted, and improve corresponding dynamic algorithms. Furthermore, we test ten datasets from UCI and design contrastive experiments to assess dynamic and classical algorithms. In terms of computational efficiency, experimental results show that the dynamic method clearly precedes the classical method for handling with dynamic attribute sets in MG-IVHFIS.& nbsp;(c) 2021 Elsevier B.V. All rights reserved.

Automated summary

This article investigates the mechanisms of dynamic updating approximations caused by the variation of attributes in a multi-granulation interval-valued hesitant fuzzy information system (MG-IVHFIS) and proposes dynamic algorithms accordingly. Experimental results demonstrate that the dynamic method clearly outperforms the classical method when dealing with dynamic attribute sets.