Analysis Of Supplier's Operational Risk Based On Logistic Regression

Market competition is heating up for businesses.While doing well in its own management,the company should strictly build a stable operation of the supply chain system,which can effectively improve its competitiveness.Those enterprises also pay more attention to the construction of a supplier evaluation system to prevent possible risks.At the same time,the enterprise has accumulated a large amount of data in the process of its own operation and the cooperation with suppliers.Therefore,how to use the available external data to evaluate the suppliers and prevent the potential business risks of suppliers from causing losses to the enterprise,under the premise of the existing internal information,is the main content of this subject.It is also where the innovation lies,that is,to provide a new method of evaluating suppliers and a reliable data sourceThe business risk model is mainly based on the logistic regression algorithm,which combines theoretical derivation with experimental analysis,to solve application problems.First,we discuss how to choose the appropriate data source,then use correlation analysis and K-S test to select indicators which are suitable for building the model.Secondly,considering that the overall sample size is small,for the negative classification samples,suppliers with large internal transaction volume and stable supply are selected as supplements,and the supplier's operating risk model is obtained based on logistic regression.Thirdly,for the class imbalance problem,that is,the number of enterprises with difficulty in business in the sample is less than another class,we use the SMOTE algorithm to increase the number of positive classification samples,and finally get an improved business risk analysis model based on logistic regression.Finally,confusion matrix and F-measure analysis are used to calculate the prediction accuracy of those two models,the ROC curve is used to compare the models.The model with higher accuracy and better generalization effect is selected,which is verified by more data.