This article presents the Early Warning System by using the financial information among 43 domestic banks from 1998 to 2002. The implementation results based on hierarchical cluster analysis, the banks can be divided into three categories: best, better and inferior in operation performance. Factor analysis got the financial variables including Earnings, Capital Adequacy, Growth, Liquidity, Asset Quality and Market Risk. Earnings is the most important, and it is significant different among three categories. Chung Shing Bank and Kaohsiung Business Bank have urgent problems to be solved. Besides, the Farmers Bank of China、Chang Hwa Com. Bank and Ta Chong Bank must be under supervision. The result of discriminate analysis is similar between Multivariate and Nonparametric, and in the latter the Normal kernel method is better than Uniform.