Enhancing Financial Credit Assessment Accuracy with Deep Learning: A Multi-Layer Perceptron Approach

Abstract

With the rapid evolution of the financial sector and the increasing complexity of credit-related data, traditional credit assessment methods often struggle to deliver accurate and generalizable results. This study explores the application of a multi-layer perceptron (MLP) network-a classic deep learning model-in the domain of financial credit evaluation. Leveraging its ability to model nonlinear relationships and process high-dimensional data, the MLP network demonstrates superior performance compared to conventional machine learning algorithms such as support vector machines (SVM), random forests (RF), and gradient boosted decision trees (GBDT). Through comparative experiments, the proposed model achieved higher precision, recall, and F1 scores, indicating its robustness in identifying complex patterns within financial datasets. Despite the model's promising performance, challenges related to interpretability, training efficiency, and handling imbalanced data persist. Future research directions include the integration of advanced techniques like transfer learning and attention mechanisms to further improve the model's adaptability and transparency. This study highlights the growing importance of deep learning in financial risk prediction and contributes to the development of intelligent, accurate, and fair credit assessment tools.