BP Neural Network Start-up Loan Based On SFLA Optimization Research On Credit Evaluation Model

With the concept of “mass entrepreneurship,innovation”,all parts of the country are providing entrepreneurial loan application channels.At present,the risk assessment of entrepreneurial loans is mainly completed through expert experience evaluation.There are certain subjective factors in the evaluation results obtained by this method,and when the number of loan applications is too large,the efficiency of approval is also not guaranteed.In view of the fact that there are many subjective factors and low audit efficiency in the current venture loan approval process,this paper mainly studies the following three aspects:Firstly,according to the subjective factors of the entrepreneurial loan approval process,based on the credit evaluation system of commercial banks,the data characteristics of entrepreneurial loans,a credit evaluation system based on entrepreneurial loans was constructed.The consistency ratio feature selection method is used to filter the redundancy indicators in the credit evaluation indicators.Refer to the commercial bank valuation method and the analytic hierarchy process to assign weights to the 19 indicators of the entrepreneurial loan credit evaluation system,thus completing the construction of the credit evaluation system.Secondly,based on the research of the technical principle of SFLA algorithm,the SFLA-3 algorithm is proposed to solve SFLA's shortcomings of slow convergence and low accuracy.The SFLA-3 algorithm uses the randomized uniform design method to improve the population initialization method to achieve uniform population distribution.The mutation operator is used to replace the random parameters in the step evolution formula,which solves the problem of blindness in step update.The method of simultaneous evolution of multiple poor fitness individuals is used to improve the method of individual evolution of the least adaptable individual to achieve the effect of improving the convergence speed.The experimental results show that the SFLA-3 algorithm is superior to the traditional SFLA algorithm in terms of optimal value accuracy and convergence speed.Finally,in view of the low efficiency of entrepreneurial loan review,this paper combines SFLA-3 algorithm with BP neural network,proposes and designs a SFLA-3_BP neural network credit evaluation model based on the credit evaluation system of entrepreneurial loan.The model calculates the optimal initial weight threshold of BP neural network by SFLA-3 algorithm,so as to accelerate the convergence speed and improve the evaluation effect.The experimental results show that the SFLA-3_BP neural network credit evaluation model has faster convergence speed and better evaluation effect than the traditional BP neural network algorithm construction model.