Assessing Credit Risk Of Cross-border E-Commerce Sellers Based On BP Neural Network

With the development of cross-border e-commerce,the entry barrier of transaction subjects are increasingly lowered,and sellers of different sizes and operating conditions participate in cross-border e-commerce transactions,and the number of transaction disputes with them is increasing.Untrustworthy behaviors such as online sales of fakes,misleading publicity,price fraud,and delayed shipments have damaged consumers' interests,inhibited their normal consumer demand,and hindered the healthy and orderly development of cross-border e-commerce.Premised on the literature review,this thesis investigates the development process of cross-border e-commerce,the relationship between cross-border e-commerce and international trade,the credit risk assessment of transaction subjects,the selection of evaluation indicators and the selection of evaluation models.Based on the credit risk assessment of cross-border e-commerce,the rational behavior theory,transaction cost theory,information asymmetry theory,and game theory are used to analyze the transaction behavior of buyers and sellers.In terms of research methods,this thesis introduces BP neural network based on machine learning and artificial intelligence into seller credit risk evaluation with a complex internal relationship,and constructs a theoretical model of seller credit risk evaluation based on the BP neural network.The basic idea of the model is to select the seller-related credit risk index as the input index of the BP neural network,and carry on the principal component analysis of these indicators,reduce the redundant variables,establish the comprehensive index,and calculate the credit risk index by entropy method as the output index of the BP neural network.Put the selected seller input index and output index as samples into the BP neural network,take part of the seller's data as training samples for machine learning and training,and take the other part of the data as prediction samples to test the fitting effect of BP neural network on credit a risk index.In the empirical part of this thesis,based on the above theoretical model,14 indicators of 64 sellers of Tmall International are selected,and the BP neural network is used for empirical analysis.The prediction accuracy can reach 99.93%.The neural network can quickly predict the credit risk value of any seller with relevant indicators,and with the increase of the sample size,the predicted results of the model will be closer and closer to the true value of the credit risk index.In this regard,credit risk managers and cross-border e-commerce platforms can use this model to detect and manage the seller's credit risk.The seller itself can also evaluate its own credit risk according to this method,and modify and improve it according to its characteristics,so as to improve its credit and business performance.However,due to the current cross-border e-commerce platform restrictions on the disclosure of seller's credit-related indicators,as well as the limitations of seller's credit risk research and application results,the obtained credit risk indicators are difficult to fully measure seller's credit risk.