The Research On Farmers' Microfinance Credit Risk Evaluation System Based On Small Sample

Farmer credit credit risk evaluation is to determine the size of the credit risk of different Farmers. Main characteristics of microfinance are small amount per loan and be in a large demand. Domestic loans with traditional subjective review are time-consuming and not conducive to risk control. It will provide decision reference that banks regulate credit standards, and boost bank the capabilities of credit risk management, proper guidance and support to the subsequent loan pricing, loan portfolio management and spection management if establishing a rational evaluation system.In this paper, delete indicators with large correlation coefficient by correlation analysis to avoid indicators reflecting duplicate information. Based on rank sum test and regression parameters significance test by expanding the samples, this paper delete indicators which can't distinct between non-default and default state. This paper established microfinance credit risk evaluation indicators system. Determine wrights of indicators with logistic regression model. This paper constructed the credit risks evaluation system of microfinance for farmers with 9 indicators by using 21% of indicators reflecting 82% of the original information finally.Main work of this paper:(1) Established the credit risks evaluation system of microfinance for farmers through correlation analysis, rank-sum test, and regression parameter significance test t. (2) Calculate and sort credit score of 2 044 farmers based on the establishment of credit risk evaluation system.The contribution charaeteristics lie on three aspects:(1) Firstly, delete indicators which can't distinct between non-default and default customer base on whether indicators have a significant distinction between non-default and default customers. (2) Secondly, through a small sample including 2 044 customers expanded to its 1 000 large sample, this paper established 1000 regression equations with non-compliance variable y and variable xi. Variable xi refers to the remained other indicators to be selected that were removed from the rank sum test. By the mean of 1 000 regression coefficients for Z test, determined indicators which can still distinguish between the default and non-default state. Without assumptions on distribution of datas and the need of increment the new sample information, it has solved the problem that how to make the the small sample into big one, and avoided credit evaluation indicators which were deleted by mistake in the small sample situation.