The Application Of Machine Learning In Credit Rating

Credit rating problem, in short, is repayment ability assessment and prediction. Although depending on the evaluation of the subject and the object of evaluation, credit rating credit rating can be divided into internal and external credit ratings (evaluation of different subject) and individual consumers’credit rating, credit rating business users (different evaluation of the object). This paper is mainly made for internal credit ratings for business users.In the introduction, mainly introduces the research background, discusses the importance of the credit rating problems. Introduce the development of the traditional credit rating, noting the importance of the application of machine learning in the field of credit ratings. And then describes two simple machine learning methods covered in this article-the artificial neural network and support vector machine concepts.In the second chapter, the paper introduces the acquisition and preliminary processing methods herein may sample data. Next is the main part of this article.In 2.2, the main use of artificial neural networks for real estate, oil and gas industry and hybrid industries three sample data modeling, forecasting and credit rating, combined with Monte-Carlo gives a more stable rate of correct, given financial indicators of relevance.In 2.3, the main use of support vector machine approach to real estate, oil and gas industry and hybrid industries three sample data modeling, forecasting and credit rating, and the neural network prediction results were compared.In 2.4, the study of the distribution of prediction accuracy of the simulation data, and distribution fitting, KS statistic compares choose a more accurate fitting of distribution and determine each prediction accuracy of the confidence interval. Furthermore the use of resampling methods used herein refers to the bootstrap method (basic bootstrap method, quantile bootstrap method, BCa bootstrap method), the average prediction accuracy rate fluctuations make analysis, confidence intervals estimated average prediction accuracy rate.In the end, the credit rating of the machine learning application results in a summary of the article, and point out the flaws and the model can improve the direction, but more important to point out the necessity of application in the credit rating of the machine learning methods.