Research On Credit Rating Of Small,Medium And Micro Enterprises Based On Support Vector Machine

Since the reform and opening up,small,medium and micro enterprises have created huge market profits for the country while possessing a huge volume.If small,medium and micro-sized enterprises want to follow the trend of The Times to expand themselves,capital is the key.However,due to the lack of data transparency of smes and miscellaneous loan risk factors,financial institutions are cautious and strict in the approval of smes' loan applications.Therefore,it is urgent to develop a fair and reasonable credit rating system and credit rating method.In recent years,premier li keqiang put forward the concept of "mass entrepreneurship and innovation".In the context of mass entrepreneurship and innovation,the key to the rapid development of small,medium and micro enterprises is to increase investment in r&d and innovation and build up their vitality.According to the current research,the credit rating system contains few innovation indicators,lacks complete indicators on innovation,and has never explored the relationship between innovation and credit rating of small,medium and micro enterprises.This paper selected four indicators to refine enterprise innovation input,the new three board manufacturing data using wind database,according to the LOGISTIC regression analysis model is established,probe into four points index to the influence of micro,small and medium enterprises credit rating,through SPSS result-2 in the logarithmic likelihood test,Hosmer and Lemeshow goodness-of-fit test,Wald test shows that after joining four refining indicators such as model fitting effect is better,higher prediction accuracy,have joined the necessity of refining indicators.In this paper,based on the classification criteria and enterprise characteristics of small,medium and micro enterprises,the selection principle of rating indicators of small,medium and micro enterprises,and the impact of innovation investment on the credit rating of small,medium and micro enterprises,four detailed indicators were added to the index system.On the basis of learning from the existing research results,we will draw up 9 primary indicators and 30 secondary indicators to build a complete credit rating index system for small,medium and micro enterprises.Compared with other machine learning algorithms,support vector machines are better at solving the problems of too small sample size,nonlinearity and too high dimension.Therefore,the SVM model was selected as the rating model in this paper and compared with the random forest and BP neural network models commonly used in the rating field.To wind database within the new three board 516 small micro enterprises as the data source in the home,first sample data is divided into training set and test set,using SMOTE algorithm equalization of sample pretreatment,the balance of positive and negative samples,proportion to 1:1,then using the learning curve to optimize the model parameters,determine the optimal parameters,the final will be trained to estimate in the test set,test results through the measure Accuracy,Specificity,F1,ROC curve,AUC area to evaluate,can be seen from the various index assessment model of excellent results.Finally,the results of the three models are compared,and the results show that the SVM model has the best effect,and the support vector machine is more applicable in the credit rating field of small,medium and micro enterprises.