The Research And Application Of Bank Loan Decision Model For SME

Since reform and opening, SME plays a significant role in the national economy andsocial development. In recent years, financing difficulties of SME has become the focus of thecommunity. On the one hand, because of imperfect financial system, weak ability against risk,information asymmetries between banks and SME and other reasons, it is difficult for SME toobtain bank loans; On the other hand, due to increasingly fierce competition among banks andconvergence in the choice of target customers, credit resources are concentrated to largecustomers. Weakened bargaining power and narrowed spreads space make the time for banksopening up new business areas and creating new profit growth. How to develop a suitablemodel for SME’s loan decision to achieve risk-benefit balance and win-win between SMEand banks, is an important issue placed in front of us.Based on the full research and analysis of financing status and risk characteristics forSME, this thesis established a bank loan decision model for SME by studying literature andrelated technologies in bank loan decision. This model has small calculated amount, strongapplication and accurate prediction. It can be effectively applied to bank loans decision forSME.This thesis has mainly completed the following researching tasks:1. Establishment of SME financial evaluation index system. This thesis conducts apreliminary screening for SME financial evaluation indicators by using financialanalysis methods. Regression model is established with the help of econometricsoftware Eviews. Through the model test, it Eventually establishes SME financialevaluation index system composed of eight core index, which is the determinantsaffecting the decision of SME bank loans.2. Combined algorithm is adopted to establish bank loan decision model for SME. This thesis collects the relevant financial data of listed companies as the sample andstandardizes. Using MATLAB as development language, it improves the parametersof the RBF neural network by genetic algorithm, and gradient descent method is usedto further optimize the results. So that the final prediction results are obtained.3. The experimental simulation and model application. Based on the trained bank loandecision model for SME, This thesis uses the RBF neural network with or withoutgenetic algorithm optimization respectively to do the simulation experiment. Resultsshow that the introduced genetic algorithm improves the forecast accuracy. Than themodel is applied to real SME bank loans, the analysis of the predicted resultsprovides the basis for loan decision.