Three-way decision-theoretic rough set model is a probabilistic extension of the algebraic rough set model.The required parameters for defining probabilistic lower and upper approximations are calculated based on cost functions through Bayesian decision procedure.Through providing different cost functions

decision-theoretic rough set model can derive many other probabilistic rough set models

such as variable precision rough set model

etc.This paper constructs an optimum problem based on decision-theoretic rough set model.Through solving the optimum problem

one can get the proper cost functions and thresholds without any preliminary knowledge.An adaptive learning parameters algorithm is also proposed to solve the optimum problem.The search space of the algorithm is the set of all instances'probabilities.Under the three-way decision-theoretic rough set model which is based on the learned cost functions and thresholds

the decision cost is minimal and a better classification performance can be gotten from that.The experimental result on some data sets shows the efficiency of our algorithm.