INTEGRATING GENERATIVE ADVERSARIAL NETWORKS AND STACKED AUTOENCODER NEURAL NETWORKS FOR ENHANCED CREDIT RISK ASSESSMENT IN FINTECH

Abstract

With the rapid development of financial technology, traditional credit risk assessment methods are struggling to cope with modern financial data challenges such as high dimensionality, sparsity of big data, and sample imbalance. This study aims to propose a new credit risk assessment model by integrating Generative Adversarial Networks (GAN) and Stacked Autoencoder Neural Networks, to overcome these challenges. Our model leverages the ability of GANs to generate realistic data samples and the advantage of Stacked Autoencoders in effectively extracting complex data features, thereby enhancing the accuracy and reliability of credit risk assessment. In the experimental part, we conducted extensive tests of the model with various parameter configurations to evaluate its performance under different conditions. The results show that our model outperforms traditional methods in key performance indicators such as accuracy, AUC value, and F1 score, especially in handling imbalanced and high-dimensional datasets. Furthermore, we provide a detailed analysis of the model’s training process, algorithm complexity, and the impact of different parameter settings on performance, to offer an in-depth understanding of how the model works. The main contribution of this study is to provide a credit risk assessment solution that combines the latest deep learning technologies, showing great potential both theoretically and practically. It not only has application value in credit risk assessment issues in the field of financial technology but also provides new directions and ideas for future related research.