BP neural network-based early warning model for financial risk of internet financial companies

Abstract

AbstractWe built an early warning model for financial risk using a back propagation neural network. To this end, the financial data of 136 listed Internet financial companies in the People’s Republic of China were selected, spanning from 2010–2019, as the sample for the empirical test. We categorized the financial status of enterprises as either “healthy” or “early warning” by the K-means clustering algorithm. Furthermore, factor analysis was performed to obtain seven common factors for building the early warning model. Overall, we confirmed the model’s excellent comprehensive accuracy and prediction efficiency, with accuracy, precision, recall, and specificity rates of 99.51%, 99.71%, 99.71%, and 98.30%, respectively. Thus, the model obtained by training and simulation using the back propagation neural network algorithm can effectively screen enterprises with hidden financial conditions and will not misclassify enterprises with good financial conditions. Notably, the misjudgment and omission rates are considerably low. The model is highly capable of identifying the financial status of Internet financial companies and has good predictive power.

Keywords

• Back propagation
• early warning
• financial
• Internet financial enterprises
• K-means clustering
• neural network