Resilience assessment of China’s strategic mineral supply chain: A hybrid GRA–BP–LSTM approach for risk prediction and early warning

Strategic mineral resources constitute the fundamental pillars of national security, industrial stability, and technological progress. In China, these resources underpin over 70 % of the nation’s GDP through extensive downstream applications. However, accelerating demand from high-tech industries—including renewable energy, aerospace, and artificial intelligence—has deepened structural vulnerabilities within the mineral supply chain, which is marked by import dependence, uneven innovation capacity, and geopolitical exposure. To evaluate these risks, this study constructs a hybrid analytical framework integrating grey relational analysis (gra), backpropagation neural networks (bp), and long short-term memory (LSTM) models. The framework combines GRA-based feature weighting with BP–LSTM’s nonlinear and temporal learning capacities to enhance predictive robustness. Using longitudinal data from 2001 to 2021, the model accurately forecasts security trends through 2025 and achieves consistent performance under fivefold cross-validation and sensitivity tests. The findings indicate that China’s strategic mineral sector remains at a “secure” level (Level IV) but continues to face challenges in R&D investment, foreign technology reliance, and international market competitiveness. A comparative evaluation with the OECD Mineral Risk Index validates the consistency of China’s GB/T 17766–2020 standard with global assessment frameworks. The study further emphasizes three policy priorities: strengthening domestic governance through resource tax reform and strategic reserves, diversifying supply networks across China–Mongolia–Russia and Southeast Asia, and advancing global coordination under OECD and International Energy Agency (IEA) mechanisms. These results provide both theoretical and practical insights for enhancing supply chain resilience and achieving high-quality, low-carbon industrial transformation.