Research On Credit Risk Evaluation Of Small And Medium - Sized Technological Enterprises Based On Spectral Clustering

Since from the reform and opening-up policy, the development of economy is fast and the industrial economy is moving towards the knowledge economy. As the technology-intensive and knowledge-intensive high-tech enterprises, the technology-based small and mid-sized enterprises are the powerful impetus for the economic development of China and become a major force for promoting economic transformation and upgrading. The amount so far has accounted for over ninety percent of the total science and technology enterprises, involved in areas such as science and technology service industry, pharmaceutical manufacturing, computer application services and so on. Ensured in the position and function of small and mid-sized enterprises, the internal financing problems of enterprise have been restricted the development of the enterprise. Small and mid-sized enterprises which is given priority to the intangible assets find it difficult to get bank credit support.As the professional technology method of the measurement and assessment of credit risk, Credit risk evaluation can help enterprises not only to guard against operational risks, to create good conditions for the modern enterprise system, but also to achieve the equity and justice of capital market. For bank, confirmation to the risks of business loans, and the management of credit assets must be established on the basis of credit risk evaluation. Since from the 1980 s, the credit risk evaluation has been introduced to our country’s capital and financial markets more than thirty years. Now gradually formed an industry, but so far it is still not enough in-depth in the understanding and specific operation of credit evaluation. Therefore, the introduction of professional evaluation method to build an effective credit risk evaluation model has become a hot issue of the academic research.Given spectral clustering algorithm can be clustering in arbitrary shape of sample space, and have the advantages of overall optimal solutions, the spectral clustering algorithm will be introduced to the credit risk evaluation of small and mid-sized enterprise. First, to establish the evaluation index system through the choice of enterprise financial data. Second, on the basis of data pretreatment of the 195 sample companies, using spectral clustering algorithm to cluster analysis, and the spectral clustering algorithm are compared with the traditional k-means clustering algorithm. On this basis, to calculate the comprehensive score of the credit risk of the technology-based enterprises through the principal component analysis, combined with the enterprise’s credit rating from the previous clustering results, so as to realize the credit risk evaluation of the small and mid-sized enterprises. Finally, according to the enterprise’s credit rating, establishing the transfer matrix and the regional transition probability matrix of enterprise is to analyze the transfer situation of the enterprise credit risk and the differences of the regional credit risks. The analysis results show that the spectral clustering algorithm can get good application in the credit risk evaluation of the small and mid-sized enterprises. The credit risk of science and technology small and medium-sized enterprise in our country has regional differences in the eastern, central and western region; Generally, the eastern region is lower than the central and western regions, the central region is lower than the western region. The model building and empirical research provide a certain reference to the credit risk evaluation of the small and mid-sized enterprises.