Font Size: a A A

Research On Calculation Models And Algorithm Optimizations Of Credit Risk Control In The Commercial Banks

Posted on:2004-12-15Degree:DoctorType:Dissertation
Country:ChinaCandidate:L M JiangFull Text:PDF
GTID:1116360125958097Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
In this paper, the author applies methods such as clustering analysis, techniques of discriminant analysis and analytical hierarchy process in building up practical credit risk control models, uses genetic algorithms and fuzzy mathematics methods to improve the general algorithms and brings these into effect by software, which bring desirable conclusion in measurement and controlling risks of loans combination, tracking forewarning monitor testing of credit risk in commercial banks, evaluation of loan profits and risks. The arrangement of this thesis is as follows:The first chapter introduces a variety of credit risk control models which applied in commercial banks of developed countries, and offers an analysis the upon characteristic and appliance scope of all kinds of models by comparison, meanwhile, recommends various calculate methods of the models. On the basis of these, the author points out that credit risk control models of commercial banks cannot be indiscriminately imitated from western countries, but appropriate to our country.The second chapter builds up a decision-making model of loans portfolio optimization based on principle of maximum earning-per risk, and analyses the meaning and characteristic of this model. This model is a type of Knapsack Problem which has restrictions between lower-limit 'and upper-limit, whose object function is the ratio between the average net present value of selected enterprise and the covariance of the average netpresent value-the net present value does not change with the numberof loans in the same direction. The model is a discontinuous and multi-maximum complicated problem. By making use of genetic algorithms with dual-structure codes, greedy algorithms, local search algorithms and optimal storage strategy,the thesis solves this complex and particular Knapsack Problem, and the convergence rates is perfect.On the grounds of corporation's life distribution and the probability of bankrupt, the third chapter puts forward a measurement and control loans risk model in commercial banks that is a linear programming problem with upper limit. To use simplex method to solve the type linear programming problem, designs a algorithm by calculating the original admissible base and admissible part to reduce restrictive conditions and decrease the scale of the matrix. The simulation results verify the availability and feasibility of the model and algorithm.The fourth chapter introduces a dynamic clustering analysis, which combines genetic algorithms and mountain climbing algorithm, improves the normal k-mean clustering analysis. This method obtains the correct results while applied in the clustering analysis of 1200 listed companies in our country. By applying Fisher multi- population discriminant analysis method, the author introduces a tracking forewarning monitor testing of credit risk in commercial banks which is fit for our facts, frames the sort standard, and institutes the method of enterprise credit evaluation grading.Taking banks as creditor, taking the enterprise's solvency as the main analysis object, according to the estimation index forced by Ministry of Finance from 2002, the fifth chapter improves the Edward I. Altaian's Z-Scole bankrupt prediction model, and builds a Z-Scole bankrupt prediction model which suits our national conditions. In the end of this chapter, the author analyses and estimates our listed companies.The sixth chapter adopts fuzzy mathematics theory and fuzzy clustering analysis methods, builds the fuzzy evaluation model of quantifying credit risk, and puts forward clustering algorithm to check outlier based on similar coefficient sum, which can raise the efficiency of checking outlier in clustering analysis. It comes to a perfect conclusion while applying this algorithm in fuzzy evaluating of quantifying credit risk. At the end, the author sets up a non-financial factors loans evaluation model, which estimates the debtor's non-financial factors combining analytical hierarchy process with fuzzy comprehensive judgment. This precise metho...
Keywords/Search Tags:credit risk control, calculation model, algorithm optimization, genetic algorithms, clustering analysis
PDF Full Text Request
Related items