Font Size: a A A

The Study Of The Private Enterprise Credit Rating Based On Support Vector Machines

Posted on:2008-12-04Degree:MasterType:Thesis
Country:ChinaCandidate:Y J PeiFull Text:PDF
GTID:2189360215952341Subject:Quantitative Economics
Abstract/Summary:PDF Full Text Request
Along with the continuous deepening of reform and improving of the socialist market economic system. China's private economy has become the most dynamic economy and the vitality bright spot, it's an important force that supports and promotes the economic growth. Three historical changes have taken place in China's private economy during the 10th"five-year plan"period, first, the status of private enterprises increasingly improved. Private Economy Blue Book "Chinese Private Economy Development Report Volume III" point out that China's private economy has become the main channel of industrial investment, employment and foreign trade. Private industrial enterprises will maintain an average annual rate of 50% in both total assets, sales revenue, the number of enterprises and industrial added value and profits. Second, the quality of private enterprises has been greatly improved during the 10th"Five-Year Plan"period, for example, the scale of the private entrepreneurs with college degrees or higher is 38.4% in 2000, it reached 51.8% in 2004. Third, the private economic policy and system reform has made substantial breakthroughs, With the spread of "36 bar of the non-public economy", a series of non-public economic development policies introduced one after another. The non-public economy system has been basically established in China.Although the state has promulgated laws and regulations to avoid the "sub-national treatment" of private economy.However,the financing of private enterprises is still facing a very big problem. Major credit rating is the current standard of large state-owned enterprises and foreign-funded enterprises. As for the relatively low-scale enterprises, private enterprises, not just credit ratings. Many private enterprises have good prospects with no financial support. This applies to the urgent need for the emergence of private enterprises credit rating. Besides information asymmetry exists between the banks and credit guarantee agencies and private enterprise, faced with a large number of credit risks. The establishment of private enterprises with characteristics of a credit rating system helps solve the conflict between the private enterprises and banks, credit guarantee institutions. This step has very important significance both in the theory and reality.Private enterprises should adhere to the correct the classification standards of credit ratings. It needs to prevent discrimination against private enterprises. Based on a systematic analysis of the Private Enterprise Credit and reference of traditional businesses, the author elected 16 indicators as a basis for the next stage of analysis .The sample is collected from Shenzhen, Shanghai Stock Exchange listed companies, choose 362 civilian-run enterprises according as the actual control human. Data provided by the Wind database of Jilin University Library. Sample coverage and dissemination of the cultural industry, including agriculture, forestry, animal husbandry, fisheries, wholesale and retail trade, information technology industry manufacturing of textiles, clothing, fur, machinery, equipment, instruments, metals, non-metallic products, petroleum, chemicals, plastics, plastics, paper, printing and comprehensive stocks.A total of 63 companies listed on the Shanghai and Shenzhen Stock Exchange homes were declared ST From January 1, 2006 to December 31, 2006; there were 28 enterprises that actual control which artificial "personal", selected 20 sample of the training randomly, the other 8 samples for testing. Meanwhile, we still have another 20 non-ST private enterprises selected as samples, ST-30 as a sample of private enterprises.To select the appropriate kernel for mapping data, this paper several kernels of the main pack for the 5 cross-certification. Use this function tests on 38 samples. Seen from the experimental process, for different the kernels, algorithm's speed, the number of support vector and classification accuracy are not the same. This paper selected Radial Basis Function as private enterprise credit ratings Support Vector Machine kernels. To determine the length of the data types of variables and indicators of the impact of the ratings, the author designed seven model。SVM1 financial data includes the traditional four indicator variables corresponding to the data for one year, that is, using 2005 data only, SVM2, SVM3, SVM4. SVM5 in SVM1 basis, adding the growth potential targets, industry indicators, enterprise strength index, 3-year growth target of one year data were used to compare these indicators of the impact of credit ratings. SVM6 SVM1 model contains indicators with the same variables. Made the timing of the data for three years (including 2003. 2004 and 2005) to compare the influence of the length of time on the data.We utilized these models to test the 40 samples with 5 cross-examinations; the accuracy rate has reached 100%. Then use these models to test 38 samples .We can see from the result that the classification accuracy of the model SVM3 has been significantly improved after industry indexes joined, it reached 86.8421%. SVM2 joined the growth potential of the model and the addition of three-year growth rate is accurate classification model SVM5 increases, which reached 84.2105%. But adding to the strength of the enterprise data model SVM4 for three years and the classification accuracy of the model and SVM6 not improved, remain at 81.5789%. SVM7 structure, it contains four types of traditional indicators, also included in the above analysis is a good indicator of the performance and growth potential, industry indicators and three-year growth targets Radial Basis kernel and still function parameters v still set up for the 0.125 and C for 2048. Test the 40 samples for 5 cross-examinations; the accuracy rate is 100%. Then we sentenced 38 training samples back, the accuracy get 89.4736% . Through this paper, the study found that for the private enterprise model credit rating system, we should not excessively consider the size indicators, but pay more attention on the development potential of private enterprise, especially the key asset and profit growth, in addition, the data model for three years didn't made a good classification results, This means that the rating for the most recent year of data is even more important.It can be seen that SVM7 best, so it has been established as credit rating model for private enterprises. The model consists of seven sub-system, they were indicators of profitability, solvency indicators, the operating capacity target capital structure targets. growth potential targets, industry classification and three-year growth target, date while is one year. Each sub-system is composed of several specific indicators to form an organic whole interconnected, there into industry Classification using qualitative indicators, the six other indicators are quantitative indicators.
Keywords/Search Tags:Enterprise
PDF Full Text Request
Related items