基于GRA-SVM的房地产上市公司信贷风险评价研究

发布时间：2019-06-11 01:50

【摘要】：房地产业作为我国国民经济的支柱产业，具有资金密集性、投资回报期长、高利润、高风险等特点。近年来房地产行业的飞速发展使商业银行对房地产开发企业的贷款比重不断增加，同时也加大了风险。因此，为降低银行的不良贷款率，保证银行资产的优良程度，对贷款的房地产企业进行有效的信贷风险评价是十分有必要的。本文站在商业银行的角度，在明确房地产信贷风险相关概念，总结信贷风险评价方法的基础上，结合房地产开发企业的特点，从房地产信贷风险产生的机理着手，找出影响信贷风险的宏观因素及微观因素。着重分析了宏观因素与房地产整体违约率的数量关系。将房地产背景实力和影响房地产行业的宏观因子等引入信贷风险评价体系，初步建立了含有29项指标的评价体系，并运用灰色关联分析法对指标体系进行约简，达到降维目的的同时也分析了这些指标对评价结果的影响程度，最终选取20个指标对房地产公司信贷风险进行评价。在对信贷风险的评价上，本文建立了支持向量机分类模型，用四种不同的核函数训练样本，运用网格搜索法和交叉验证法进行核参数寻优，建立支持向量机，，再用得到的不同决策模型对测试样本进行测试，比较四种核函数的分类结果发现，径向基核函数的分类准确率最高，建立模型难度也较低，性能最优。通过与单一SVM方法、Logistic回归分析方法进行比较，结果表明基于本文提出的GRA-SVM方法的分类准确性和推广能力明显好于其它几种方法，证实了该方法的有效性和可行性，为商业银行银行建立可靠的房地产公司信贷风险评价系统提供了依据。
[Abstract]:As the pillar industry of our national economy, the real estate industry has the characteristics of capital intensity, long return period of investment, high profit, high risk and so on. In recent years, with the rapid development of real estate industry, commercial banks have increased the proportion of loans to real estate development enterprises, but also increased the risk. Therefore, in order to reduce the non-performing loan ratio of banks and ensure the excellent degree of bank assets, it is very necessary to evaluate the credit risk of real estate enterprises. This paper stands from the point of view of commercial banks, on the basis of defining the related concepts of real estate credit risk, summing up the credit risk evaluation methods, combining with the characteristics of real estate development enterprises, starting from the mechanism of real estate credit risk. Find out the macro and micro factors that affect the credit risk. This paper focuses on the quantitative relationship between macro factors and the overall default rate of real estate. The background strength of real estate and the macro factors affecting the real estate industry are introduced into the credit risk evaluation system, and the evaluation system containing 29 indexes is initially established, and the index system is reduced by grey relational analysis. At the same time, the influence of these indexes on the evaluation results is analyzed, and 20 indexes are selected to evaluate the credit risk of real estate companies. In the evaluation of credit risk, this paper establishes a support vector machine classification model, uses four different kernel functions to train samples, uses grid search method and cross verification method to optimize the kernel parameters, and establishes the support vector machine. Then the test samples are tested with different decision models. The classification results of the four kernel functions show that the classification accuracy of the radial basis kernel function is the highest, the difficulty of establishing the model is low, and the performance is the best. Compared with the single SVM method and Logistic regression analysis method, the results show that the classification accuracy and generalization ability of the GRA-SVM method proposed in this paper are obviously better than those of other methods, and the effectiveness and feasibility of the method are verified. It provides a basis for commercial banks to establish a reliable credit risk assessment system for real estate companies.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.45;F299.23;F224

【参考文献】