This thesis establishes an information system which combines three credit risk models through Service-Oriented Architecture (SOA). The system requires the bank user inputting finance-related data and selecting options to generate a series of credit risk related results, including the probabilities of default, the recovery rates, the expected market value of assets, the volatilities of the expected market value of assets, the default points, the default distances, and four indexes from principal components analyses. In addition to exhibiting the numerical results, graphical results are also available for the user. Three credit risk models joining this system are the Moody¡¦s KMV Model with Default Point Modified, the Risk-Neutral Probability Measure Model, and the Time-Varying Jointly Estimated Model. Several previous researches have demonstrated the validity of these credit risk models, hence the purpose of this study is not to examine the practicability of these models, but to see if these models are capable of connecting each other effectively and eventually establishing a process to evaluate the credit risk of enterprises and industries by the use of testing samples. Testing samples are data from Taiwan Small and Medium Enterprise Credit Guarantee Fund. The finance-related data includes the loan amounts, the book value of assets, the data used to calculate the default point threshold (such as the short-term debt and the long-term debt), and the financial ratios with regard to growth ability (such as the revenue growth rate and the profit growth rate before tax), operation ability (such as the accounts receivable turnover rate and the inventory turnover rate), liability-paying ability (such as the current ratio and the debt ratio), and profitability (such as the return on assets and the return on equity). In addition to inputting the finance-related data, the system also require the user selecting the industrial category, the default point threshold, the way data being weighted, the data period, and the borrowing rates from the option page for every enterprise in order to acquire the results. Among the computing process, user is required to select weighted average method, either weighted by loan amounts or weighted by market value of assets, to obtain ¡§the weighted average probability of default of the industry¡¨ and ¡§the weighted average recovery rate of the industry¡¨ which are both used by the Time-Varying Jointly Estimated Model. This study also makes use of quartiles to simulate the situation when the user is near the bottom and top of the business cycle. Furthermore, the ¡§Supremum Strategy¡¨ and the ¡§Infimum Strategy¡¨ are added to this study to let the user realize the best condition and the worse condition of the ¡§Time-Varying Industrial Marginal Probabilities of Default¡¨.