Approaches to attribute reductions based on rough set and matrix computation in inconsistent ordered information systems

摘要

In order to conduct classification analysis in inconsistent ordered information systems, notions on possi ble and compatible distribution reductions are proposed in this paper. The judgement theorems and dis cernibility matrices associated with the two reductions are examined, from which we can obtain an approach to the two reductions in rough set theory. Furthermore, the dominance matrix, possible and compatible decision distribution matrices are also considered for approach to these two forms of reduc tions in inconsistent ordered information systems. Algorithms of matrix computation for possible and compatible distribution reductions are constructed, by which we can provide another efficient approach to these two forms of distribution reductions. To interpret and help understand the algorithm, an exper imental computing program is designed and two cases are employed as case study. Results of the small scale case are calculated and compared by the discernibility matrix and the matrix computation to verify the new method we study in this paper. The large-scale case are calculated by the experimental comput ing program and validated by the definition of the reductions.