IRisk Assessment Model in Enterprise Investment and Financing Decision-making Based on Time Series Analysis

Abstract

In enterprise investment and financing, accurate risk assessment is crucial but remains challenging due to the complexity and high dimensionality of financial data, as well as the need for both predictive accuracy and interpretability. Existing methods, such as traditional statistical models and machine learning algorithms, often struggle to achieve a balance between these two aspects, limiting their practical applicability. To address these issues, this study proposes a hybrid risk assessment model that integrates Multi-Layer Perceptrons (MLPs) for nonlinear feature extraction with decision trees for interpretable rule-based classification. The model leverages the strengths of both approaches, offering high prediction accuracy while maintaining clear decision-making logic. Furthermore, a novel fusion strategy is introduced to enhance the synergy between the two modules. Experimental results demonstrate that the proposed model outperforms several classical methods, including XGBoost, LightGBM, and Random Forest, in terms of accuracy, F1 score, and AUC. Ablation studies validate the independent contributions of each module, as well as the effectiveness of the fusion strategy, confirming the robustness and design rationale of the model. The findings highlight the model's adaptability to varying data conditions and its ability to deliver reliable risk assessments even in complex financial scenarios. This study contributes to the field by providing a solution that bridges the gap between interpretability and predictive performance in enterprise-level risk assessment. The proposed model offers a promising framework for enhancing decision-making processes in investment and financing, laying the foundation for future advancements in data-driven financial risk management.