Research On Personal Credit Evaluation Model And Algorithm Based On Improved PSO Algorithm And BP Neural Network

The purpose of this paper is to build a personal credit evaluation model based on neural networks,to help commercial banks or companies establish an automated credit audit system which reduce the consumption of human resources by using computer software to calculate the lender's credit status and quantitatively evaluate the lender's credit so as to improve the degree of automation of personal credit evaluation.The main research contents in this thesis are as follows:1.Discuss the basic concepts of credit evaluation and the principles to be followed in practical work,study the core elements affecting individual credit ratings and the basic methods and models for personal credit evaluation.Then based on the German open source personal credit database,the original data preprocessing,calculating the gain of each index and sorting according to the gain value,extracting the top 17 credit indicators,and constructing the credit evaluation index system needed for the research model of this paper.2.Explore the artificial neural network,elaborate its basic concepts and principles,and thoroughly study the BP neural network and its training algorithm: gradient descent algorithm.The principle of the training algorithm is deduced and the performance and defects of the training algorithm are analyzed.In view of the flaws in the algorithm,this paper describes seven existing improved algorithms,and designs simulation experiments to compares and analyzes the performance of these seven training algorithms in the personal credit evaluation model.Finally,a personal credit evaluation model based on BP neural network is constructed..3.This paper makes an in-depth study of the particle swarm optimization algorithm and its improvement.The purpose of this paper is to use Particle Swarm Optimization(PSO)algorithm to replace the gradient descending algorithm in BP neural network to train the weights and thresholds of neural networks,so that the neural network model has better weights and thresholds.At the same time,this paper improves on the basis of the standard particle swarm optimization algorithm.The standard particle swarm optimization algorithm will narrow the search space when the shared population search wisdom,and the global search ability will decrease.The basic improvement idea based on the particle swarm optimization algorithm of this paper is to give new particles adding a dynamic random position transformation enables the global optimization of particles to be enhanced.This paper compares the improved algorithm with the standard algorithm and verifies that the improved particle swarm optimization algorithm does have better global search ability than the standard particle swarm optimization algorithm.4.Finally,this paper replaces the training algorithm in BP neural network with the improved particle swarm optimization algorithm,constructs the BP neural network algorithm model of improved particle swarm optimization,and finally realizes the personal credit evaluation model based on the improved PSO-BP neural network.A simulation experiment is designed to compare the performance of the three algorithm models BP model,standard PSO-BP model and improved PSO-BP model applied to personal credit evaluation one by one.The experimental results show that the improved PSO-BP model has the highest prediction accuracy and performance.