Three relaxation models (VC-DRSA, VP-DRSA and ISVP-DRSA) of DRSA have been proposed to relax the strict dominance principle. However, the classification performance of these models is affected by the value of consistency level
. Until now, the value of
is set according to prior domain knowledge. But no one knows which value is the best and the reason. To address the multicriteria classification problem, we propose a new method in this paper. A new uncertainty measure is defined and an algorithm for transforming inconsistent preference-ordered systems into consistent ones (TIPStoC) is designed in this paper. An iterative approach is adopted in TIPStoC algorithm. We find that inconsistent preference-ordered information systems can be transformed into consistent systems with low computation complexity, and without losing useful information. The classification performance will be improved with the decision rules induced from the consistent systems. Besides, the value of consistency level
is set to 1.0 without depending on prior knowledge. Finally, the procedure of TIPStoC algorithm is illustrated by a real example and the efficiency of the new method is proved by experiments.