Rough set theory plays an important role in knowledge discovery, but cannot deal with continuous attributes, thus discretization is a problem which we cannot neglect. And discretization of decision systems in rough set theory has some particular characteristics. Consistency must be satisfied and cuts for discretization is expected to be as small as possible. Consistent and minimal discretization problem is NP-complete. In this paper, an immune algorithm for the problem is proposed. The correctness and effectiveness were shown in experiments. The discretization method presented in this paper can also be used as a data pre- treating step for other symbolic knowledge discovery or machine learning methods other than rough set theory. Rough set theory plays an important role in knowledge discovery, but can not deal with continuous attributes, thus discretization is a problem which we can not neglect. And discretization of decision systems in rough set theory has some particular characteristics. Consistency must be satisfied and cuts for discretization The proofness and effectiveness were shown in experiments. The discretization method presented in this paper can also be used as a data pre- treating step for other symbolic knowledge discovery or machine learning methods other than rough set theory.