Research On Credit Evaluation Of Farmers In Financial

In the rural financial market, the agriculture-related financial institutions can’tscreen out the farmers with the high degree of credit timely and effectively due to thepoor credit information environment, or even they can, but the cost of screening is high.Therefore, the rural financial institutions are reluctant to loan to the rural farmers,resulting in the presence of the rural financial "paradox". The credit environment of therural is seriously lagging behind and the majority of farmers doesn’t enjoy the benefitsbecause they can’t get loans because of the lack of effective collateral.The establishment of farmers credit evaluation model can predict the credit of thefarmers accurately by selected farmers credit evaluation index, and the agriculture-related financial institutions can filter out farmers with high credit effectively with lesscosts. On one hand, it can not only improve the enthusiasm of loaning to farmers of theagriculture-related financial institutions with higher level of operations and the revenueand the lower risk. On the other hand, it can effectively solve the problems that ruralfarmers can’t get the loans, because the customers with high credit can obtain loansfinancing effectively.Firstly, this article give an overview of the theoretical study of foreign scholars onrural financial status, a summary of the successful experience of the Grameen Bank ofBangladesh and an overview on the research of the current situation, reasons andcountermeasures of China’s rural financial problems of the domestic scholars.Secondly,this article describe the application of probabilistic neural network and support vectormachine model on how to evaluat the credit of the farmers and the advantages anddisadvantages of these models.Thirdly, this article establish the basic indicators thatimpacting the credit of the rural farmers of the two provinces by analysing thecharacteristics of the rural farmers on the basic situation, basic economic activities andfinancial behaviors in detail by the use of the survey data on the rural financial of Jilinand Liaoning Province.Finally, this article filter out the high correlation indicators from the basic indicators by principal component analysis and select six common factorrepresents the information of the surveyed farmers and make them as the input of theprobabilistic neural networks and support vector machine model to train, predict andoutput the results.The results show that the basic credit evaluation index of the rural farmers hasbeen selected can effectively recognise the credit of the farmers. The probabilisticneural network model is100%correct to recognise the training samples, the predictionaccuracy rate of farmers with good credit is76.47%, the prediction accuracy rate offarmers with bad credit is70%. Though the support vector machine model is100%correct to recognise the training samples, the prediction accuracy rate of the predictionsamples and farmers with good and bad credit is77.78%as the same, and it’s betterthan the probabilistic neural network model. Probabilistic neural network model has ahigher prediction accuracy rate for customers with good credit and the support vectormachine model has a higher prediction accuracy rate for customers with bad credit.We conclude that the probabilistic neural network model and support vectormachine model can be used for the rural farmers’ credit recognition in Jilin andLiaoning Provinces and because the support vector machine model can be promotedwidely and has a better prediction accuracy rate on the prediction samples, it is moresuitable for practical applications.