Credit Risk Classification and Prediction Based on Deep Neural Network Algorithm

This study predicts whether a user has defaulted based on correlation analysis and deep neural network algorithms. The results of the study show that the occurrence of default by a user is positively correlated with age, family, years of employment, and credit length, and negatively correlated with income, amount, rate, status, and percentage of income. After model training and testing, the prediction accuracy was 81.68% on the training set and 81.68% on the test set. Specifically, there were 2858 correct predictions and 641 incorrect predictions in the training set, of which 469 incorrectly predicted that no default had occurred as having occurred and 172 incorrectly predicted that a default had occurred as not having occurred, and there were also 2858 correct predictions and 641 incorrect predictions in the test set. The results of this study show that the established model has high reliability and accuracy in accurately predicting whether a user has defaulted or not, which provides an important reference for risk assessment and decision-making.