Bank Credit Risk Analysis Based on Network Data Mining and Pre-training-fine-tuning ANN

At present, machine learning model is widely used in bank credit risk prediction, but there are still some problems in the actual use. Aiming at the limitations of single data source, static data and little data, we optimize the artificial neural network model. First, we use the network data mining technology and introduce the real-time news text data from the network as a dynamic supplement to the financial index data; The second is to use pre-training and fine-tuning strategy. Finally, we take 48 listed companies in agriculture, forestry, fishery and animal husbandry as the research objects for empirical analysis. By comparing the prediction accuracy and stability of the optimized model with that of the original model, we conclude that the optimized model has better precision improvement effect, higher data prediction stability and, more importantly, more outstanding performance in the prediction of nonperforming loans.