This paper analyzes the current operating situation of financial holding companies in Taiwan. After referring to the operation of financial early warning systems of various countries, the study chooses appropriate financial ratios to establish a financial early warning model for quantitative analysis, evaluate the management efficiency of financial holding companies, discriminate the correct classification rate of prediction probability and rating system, and seek an optimal early warning model as the basis for supervision and governance of financial holding companies. In 2008, the financial tsunami that swept over the global economy resulted in a disastrous loss to the financial industry. To cope with the impact of financial crisis, most countries in the world have developed their early warning models. In Taiwan, the CAMELS framework was adopted for the establishment of Taiwan¡¦s financial early warning system and a risk-oriented auditing system. With the financial liberalization, the government of Taiwan lifted the ban on the business operation of financial holding companies step by step in order to enhance the operating efficiency of financial holding companies and activate the financial market. However, the competitive ability of Taiwan¡¦s financial industry was not significantly improved. Instead, a series of problems with the financial sector erupted one after another. The reasons for such a condition were due to more risks faced by the financial holding companies after financial deregulation. Therefore, this study used 14 financial holding companies in Taiwan as of 2006 as subjects, and constructed a financial early warning system for the original samples by using the following two kinds of models. After factor analysis¡Athe remaining financial variables ¡Alike capital adequary ratio(C2) ¡Atotal debt/equit capital (C3) ¡A total deposit/equit capital(C4), ratio of non-performing loans(A2) the operational expense ratio(M3), efficiency ratio (M4), earnings before taxes/sales(E1) and so on, have more influence on the performances of the financial holding companies in Taiwan. As to the whole efficiency of the self-examination, CAMELS still has good prediction ability and can enable predicting ability increases after joining the risk parameters¡D Predictive sample enters two models and obtains¡Gthe predictive efficiency, type error and type error of Model Two work better than Model One¡Aso in predicting samples, think CAMELS still has good predicition ability and can enable predicting ability increases after joining the risk parameters.