Research On Combined Model Of Farmers Credit Rating Based On Logistic-SVM

also currently undergoing tremendous changes, in order to scale-based farming,aquaculture and forestry nectar rapid development in rural areas, the scale of operationsbrought the farmers considerable income. But the majority of households rely on the ruralfinancial system, mainly the rural credit cooperatives and other microfinance institutionsto maintain operations. Therefore, Rural Microfinance has a huge demand, but the smallamount of each loan, the large number of loans, is not conducive to agency management.Since the development of the microfinance started late, CDM credit institutions are notperfect and unique farmer credit risk, there will be no comprehensive credit evaluationindex, credit rating standards are not one, as well as traditional human too subjective toreview issues such as human and material cost while not conducive to risk control, in theprocess of credit assessment on farmers. Therefore, the distinction of farmers applying forloans between the qualified customers (high credit rating) and general customer (creditrating low) to control credit risk, achieve sustainable and healthy development of creditinstitutions while improving the credit evaluation mechanism of the countryside, it is tothe interests of all the stakeholders. Therefore, the establishing of an effective creditevaluation for farmers has important practical significance, which is the purpose of thispaper.This paper used theory and empirical analysis, combined with the results of previousstudies and the data characteristics of rural credit cooperatives of farmers in Jilin Provinceto establish the farmer credit evaluation index system. The system includes: the targetlevel (level one indicators) contains two categories,"good customer" or "bad customer";criteria level (level two indicators) includes four categories, namely household naturalfeatures, solvency, operating conditions and creditworthiness; index level (level threeindicators) is attached specific indicators on criteria level, the paper there were13specificindicators. Among them, the family natural features including age, education, marital status, health status, family size, and number of family members and other dependents oflabor indicators; solvency of external liabilities, including family liability, the presentvalue of housing, household income, household consumption and total assets and so onoperation conditions index for the land acreage representatives; creditworthiness mainlyrefers to whether there is a guarantor of this indicator. This paper performed empiricalanalysis, first the author processed the basic data, namely, data filtering, sample allocationand standardization, then the author got nine common factors by factor analysis, factorssuch as the evaluation of public representatives of farmers credit information is enteredLogistic model and SVM model. Then the author used the Logistic model and SVMmodel for empirical analysis and make predictions. The combined model is based on theLogistic model and SVM model by the method of least squares method to solve the weightcoefficients, modeled after the completion of its examination. Finally, the three models topredict the results of the samples were compared to reach a conclusion.The empirical results show that, Logistic model and SVM model can effectivelyevaluate the credit for farmers, but the combined model based on the two modelsperformed better in evaluating the effect of the credit for farmers. When testing samples tomake predictions, Logistic model prediction accuracy rate is79.08%, SVM modelprediction accuracy rate is75.56%, a combination of model predictive accuracy rate is84.06percent. The prediction accuracy rate of the combined model is better than each ofthe two single models, and the prediction error is less than two single models. To furthervalidate the predictive model is better than the effect of two single models, the author usedthree models to predict the test sample2analysis, the forecasting accuracy Logistic model,SVM model and combined model was71.42%,73.02%and77.78%respectively. Thisprediction proved once again that the combination model inherits two single modelpredicts with high accuracy, good stability, based on and improves both performanceindicators. This study shows that, Logistic model and SVM model can be used to evaluatecredit of Jilin farmer, but a combination of both to establish a model has higher predictionaccuracy and stability, and can better predict its effect. So, Logistic-SVM combined modelfor credit rating of Jilin farmers is more in line with the actual situation.