Research on Supplier Credit Evaluation Based on Data Mining

Abstract

Data mining method is used to clean the real data of suppliers on cost link. With its 80% random drawing as a training set, a WOE and logistic regression model is set up to distinguish the integrity supplier on the platform according to the new calculation formula of WOE for indicator variable. And using the rest 20% as test data to calculate the accuracy of this model which is as high as 96.9%. Finally, the credit score card is established according to the linear scaling relation between the WOE value and regression coefficient of each indicator variable, which can calculate supplier credit score and provide decision-making reference for the platform and users when selecting suppliers.