End-to-end supply chain resilience management using deep learning, survival analysis, and explainable artificial intelligence
研究提出一个数据驱动的框架，利用深度学习、生存分析和可解释人工智能，从企业内部运营数据预测供应链中断风险，并在美国汽车制造商案例中实现半年内关键部件短缺预测误差低于20天，错误率比最优模型降低50%。
This study introduces a data-centric framework for end-to-end supply chain resilience management. With major disruptions such as pandemics profoundly affecting industries and regions, a wealth of data capturing diverse disruption scenarios has emerged. This presents an opportunity to correlate deviations in organizational operations with disruption outcomes, reducing reliance on external supplier data and alleviating associated data privacy concerns. Utilizing deep learning, survival analysis, and explainable artificial intelligence, the research represents a pioneering advancement in translating readily accessible organizational data into forecasts of disruption risks and sources, differing from traditional model-centric methodologies. The application of this framework to a real-world scenario based on a U.S. automotive manufacturer resulted in accurately predicting the time-to-survive for critical parts, with a prediction error of under 20 days for half-year-ahead shortage forecasts. Notably, the model achieved a 50% reduction in error rates for near-term and long-term predictions compared to the best-performing alternative models. Our findings underscore the framework's ability to effectively address the complexities of global supply chain disruptions and unknown-unknown uncertainties by harnessing insights gleaned from internal operational data. This accumulated knowledge enables real-time risk identification and assessment, empowering organizations to deploy timely and targeted risk mitigation strategies for enhancing overall supply chain resilience.