Dynamic Computing Rough Approximations for Variable Granular Structure Lattice-Valued Decision Systems

A lattice-valued decision information system has condition attributes consisting real-valued, set-valued, interval-valued, fuzzy-valued, intuitionistic fuzzy-valued attribute and so on. Meanwhile, the information granule structure of information system may vary over time when new information arrives and redundant data leaves. How to quickly update the approximations of a concept invariable granular structure system which caused by adding or deleting attributes? In this paper, we propose two dynamic obtaining rough approximations approach for inserting and removing attributes, respectively. The novel updating mechanism enables additional knowledge to be obtained from the alterant datasets without neglecting the prior knowledge. Furthermore, a case study is conducted toverify the feasibility and effectiveness of the dynamic computing approaches.