Research on Supply Chain Disruption Risk Prediction Model Based on Machine Learning

Abstract

Supply chain disruptions feature sudden occurrence, transmissibility and severe losses. Traditional risk management approaches relying on manual scoring and post-incident response fail to identify complex nonlinear signals in a timely manner. Targeting the demand for predictive decision-making in operations and supply chain management, this paper constructs a machine learning prediction framework consisting of data governance, feature engineering, model training, risk interpretation and response linkage, and compares Logistic regression, support vector machine, random forest and gradient boosting tree. The study demonstrates the model evaluation process using reproducible exemplary simulation data, and measures prediction quality comprehensively through accuracy, recall, F1-score and area under the receiver operating characteristic curve. The study holds that the value of machine learning does not lie in the mere pursuit of algorithmic accuracy, but in converting risk probabilities into operational actions such as inventory adjustment, multi-source procurement, transportation rerouting and supplier collaboration. The implementation of the model should be accompanied by the simultaneous establishment of data standards, threshold governance, interpretability mechanisms and rolling update systems. This framework can effectively integrate professional capabilities including supply chain management, risk statistical modeling, predictive analytics and AI-enhanced business analytics, providing actionable research insights for supply chain resilience building.