Prediction of default risk borne by financial enterprises in China based on big data analysis approach

Abstract With the development of digital technology, enterprises' online financial business capabilities continue to improve. Predicting the default risk of borrowers is an important issue in the development of financial enterprises in China. In this study, big data analysis algorithms were used to predict the credit risk of loans. Data were collected from a large financial enterprise in China, including demographic characteristics and characteristics of the loaning process. The overall accuracy and single precision of the forecast are evaluated by cost matrix analysis. For continuous input and discrete output data pairs, decision tree (DT) model has achieved best accuracy and negative precision; artificial neural networks (ANN) algorithm has performed brilliantly in sensitivity; support vector machine (SVM) model has achieved the highest area under curve (AUC). For discrete input-output data pairs, DT model has achieved the best accuracy, sensitivity and negative precision. DT has been employed to analyze the relationship between the default risk and selected attributes, the possible causes and appropriate measures are discussed in this study.