How to avoid the bad loans and ensure the quality of creditor’s rights, is a research topic of great value in banking. Credit decision mechanism is the most basic and most important in commercial bank. Therefore, mining the potential rules from historical database to guide the current credit decision is very important.This paper first reviews the development of the decision mechanism, and points out that it is quite subjective in decision-making process. Second, we summarize the origin of data mining and the common data mining technology, and then focus on rough set theory. Third, we establish the static credit decision model which can analyze the history cases of credit decision by rough set method and reveal the main factors which affect the success or failure of credit decision and calculate the importance of attributes objectively. Fourth, as real world database is very huge, to find the optimal reduction is a NP hard problem, so we turn to seek satisfactory solution based on information entropy. In order to update the entropy, we propose the incremental algorithm and give its time complexity. The innovation of this paper lies in establishing credit decision model by rough set, and proposing incremental algorithm based on complementary entropy for dynamic information systems. The model and algorithm proposed in this paper are useful for the improvement of bank credit management system. |