Research On Cradit Risk Evaluation Based On GRA-SVM In Real Estate Listed Companys

As the pillar industry of the national economy, the real estate industry has the characteristics of capital-intensive, long period of return on investment, high risk, etc. In recent years, the rapid development of real estate industry leads to the increasing proportion of commercial bank loans to the real estate development enterprises, but also increases the risk. Therefore, efficient credit evaluation on real estate customers is the essential means of reducing bad loans possibilities and self-risk for commercial banks.The thesis bases on the perspective of commercial Banks, on the basis of making credit risk related concepts clear, summarizing the credit risk evaluation method, proceed from the generating mechanism of real estate credit risk, to identify the macro factors and micro factors of credit risk. Introduce background strength of the real estate and the macro factors affecting the real estate industry credit risk evaluation system, establish initial evaluation system containing29indicators, and carried out an index reduction using gray relational analysis, achieved the purpose of dimensionality reduction as well as analyzed the impact of these indicators on the evaluation results, and finally selected20indicators to evaluate the credit risk of the real estate company.In the evaluation of credit risk, the thesis established a support vector machine classification model, used four different kernel functions to train samples, using the grid search method and cross validation to carry out kernel parameter optimization, to establish the support vector machine, and then get different decision-making model for testing the test samples to compare the classification results of four kernel functions, then found RBF kernel function made the highest classification accuracy, and have lower difficulty as well as optimal performance.Compare with the single SVM method, Logistic regression analysis method, the results show that the GRA-SVM method based on the proposed classification accuracy and generalization ability was significantly better than several other methods, confirming the validity and feasibility of the method and provides a method for thecommercial banks to establish a reliable real estate company credit risk evaluation system.