This study proposes a prediction model based on a dynamic spatiotemporal causal graph neural network to address the challenges of complex dynamic dependencies, strong structural correlations, and ambiguous causal relationships in corporate revenue forecasting. The model constructs a time-varying enterprise association graph, where enterprises are represented as nodes and industry or supply chain relationships as edges. A graph convolutional network is used to extract structural dependency features, while a gated recurrent unit captures temporal evolution patterns, achieving joint modeling of structural and temporal features. On this basis, a causal reasoning mechanism is introduced to model and adjust potential influence paths among enterprises. A learnable causal weight matrix is used to describe the strength of economic transmission, suppress spurious correlations, and strengthen key causal paths. The model also employs multi-scale temporal aggregation and attention fusion mechanisms to dynamically integrate multidimensional information, enhancing adaptability to both long-term trends and short-term fluctuations. Experimental results show that the proposed model outperforms mainstream methods in multiple metrics, including MSE, MAE, MAPE, and RMAE, verifying its effectiveness in capturing corporate revenue dynamics, modeling economic causal dependencies, and improving prediction accuracy. This study establishes a unified framework that integrates spatiotemporal dependency modeling with causal structure reasoning, providing new insights and methodological foundations for intelligent forecasting in complex economic systems.

This paper addresses the challenges in traditional enterprise performance analysis, including complex multi-source data structures, ambiguous indicator correlations, and poor decision interpretability. It proposes an enterprise performance optimization decision model that integrates knowledge graphs with causal inference. The model constructs a multi-entity and multi-relation knowledge graph to semantically integrate heterogeneous information from financial, market, and operational dimensions, enabling high-level representation of structured relationships among enterprise features. It further incorporates causal structure learning and inference mechanisms to identify key performance drivers and estimate intervention effects, revealing the true causal pathways among variables. In the optimization layer, the model combines knowledge representation with causal relationships to establish an interpretable decision objective function, ensuring that predictions possess both numerical accuracy and causal consistency with logical traceability. Experiments conducted on public enterprise datasets demonstrate that the proposed method outperforms mainstream deep learning and sequence modeling approaches in terms of error control and generalization performance, showing higher robustness and stability. Sensitivity analysis further confirms that the model maintains strong adaptability and consistent performance under different embedding dimensions, noise levels, and optimization strategies. This study provides a novel methodological framework and theoretical foundation for building interpretable and intervention-oriented intelligent decision systems, offering significant implications for data-driven performance evaluation and decision optimization.

This study proposes a deep learning framework that combines graph neural networks (GNN) and temporal modeling to enhance the accuracy and stability of supply chain risk prediction and optimization in pharmaceutical enterprises. By modeling the pharmaceutical supply chain as a complex graph structure, this research effectively captures the dependencies between nodes while using temporal networks to capture long-term dynamic changes within the supply chain. We design a model incorporating a multi-head attention mechanism, which provides accurate risk predictions under different demand fluctuation scenarios. The experimental results demonstrate that the proposed model outperforms existing traditional machine learning models and deep learning methods across multiple evaluation metrics, including Precision, Recall, F1-Score, and AUC-ROC. Particularly in complex environments, the model effectively identifies potential supply chain risk events, such as logistics delays, supply disruptions, and inventory fluctuations. Compared to traditional rule-based or statistical supply chain risk prediction methods, the proposed model shows greater robustness and accuracy by deeply exploring the structural and temporal features between supply chain nodes. Sensitivity analysis of model performance under varying demand fluctuation intensities and environmental changes further validates the model's feasibility and stability in real-world applications, providing effective technical support for the pharmaceutical industry in areas such as resource scheduling, inventory management, and risk early warning.