The models for credit risk assessment based on certain method have been discussed in previous papers. These models could not fully depict or predict credit risk of financial institutions due to the restriction of sample amount. Based on the analysis of data cube and a sample of 1333 loans of several domestic commercial banks, this paper discloses the basic characters of credit risk, sets up two-period of credit risk assessment system. The empirical result shows that the model is stable and efficiency and it provides powerful supports for the financial institutions to build credit risk assessment system.