This study explores integrating big data and advanced deep learning techniques for enhancing personal credit risk assessment in commercial banks. Traditional methods must be improved in high-dimensional, sparse, and noisy big data environments. Key challenges include data source diversity, variable selection complexity, and methodological differences in modeling. By leveraging deep learning approaches like Stack Denoising Autoencoder Neural Networks (SDAE-NN) and addressing imbalanced data using Generative Adversarial Networks (GANs), this research aims to develop robust frameworks that improve the accuracy and efficiency of credit risk evaluation.