Research on Innovative Financing Models and Credit Risk Evaluation for SMEs in Digital Finance: Integrating Time Series and Machine Learning Algorithms

The credit risk assessment and financing model optimization of small and micro enterprises in the digital financial environment have always been important topics of concern in academia and practice. Traditional credit risk assessment methods rely on static data and are unable to fully capture the dynamic changes in corporate credit status, resulting in inefficient risk identification and inability to effectively support the financing needs of small and micro enterprises. To solve this problem, this paper proposes a credit risk assessment model based on the fusion of long short-term memory network (LSTM) and random forest (RF). By combining the dynamic feature extraction capabilities of deep learning and the feature optimization capabilities of machine learning, it improves Accuracy and robustness of credit risk assessment.The experimental design is based on real public digital financial platform data sets, and evaluates the performance of Baseline, LSTM, RF and fusion models. The results show that the fusion model is significantly better than other models in terms of accuracy, F1 score, AUC and other indicators, especially in dynamic credit risk assessment. In addition, the optimized financing model based on this model has significant advantages in improving financing success rate, reducing non-performing loan rate and improving resource allocation efficiency, verifying the practical application value of the model in supporting financing decisions of small and micro enterprises.