Abstract: Facility rating aims to evaluate the possibility of repayment or the loss given default (LGD) of one debt. It is hard for commercial banks to accurately evaluate the credit risk of small business loans, because of mall enterprises' high risk, small amount, and untrue financial data, etc. This paper established facility rating index system for small enterprises, which selected indicators based on the ability of distinguishing default state for the first time, while the second time was to avoid the information redundancy of selected indicators in the same guidelines layer. First,establishing logistic regression equation about default state and one indicator, then constructed χ²-statistic by comparing the log-likelihood values between the zero-model which has no any indicator and the full-model which has an indicator. The greater the deviation of those two log-likelihood values, the easier the indicator can distinguish the default state, that is to say, the indicator should be retained. It compensates the disadvantage that the existing research has nothing to do with the default state when screening indicators. Second,the paper avoided information redundant by calculating the correlation coefficient between any two indicators in the same criteria layer, if the correlation coefficient between these two indicators is greater than a threshold, then remove the indicator which has small χ²-value. This method can avoid redundant information and mistakenly deleting the indicator which has more significant impact on default state. What's more, it changes the disadvantage that the existing researches subjectively delete one indicator when the correlation coefficient is greater than the threshold. The results shows that the established debt rating system of small industrial businesses, including 26 indicators, such as asset-liability ratio, cost margins, the corporate credit situation nearly 3 years, by extracting the related data of 28 regional commercial bank branches of China. the paper can not only get the credit rating of the loan enterprises, but also get the loss given default of per credit rating. It changes the disadvantages that the existing research can only calculate the credit score and give credit rating.

Key words: small industrial enterprises, credit risk, debt rating, default state, Likelihood ratio test