Hybrid GNN-LSTM Model for Real-Time Supply Chain Risk Prediction

Demand forecasting, inventory optimization, and risk mitigation are fundamental and crucial parts of Supply Chain Management. Recent advances in Machine Learning and Deep Learning have brought into existence some promising opportunities for predictive analytics in complex supply networks. A hybrid model is proposed in this study, combining Graph Neural Networks (GNNs) and Long Short-Term Memory (LSTM) networks to forecast supply chain disruptions by learning spatial (i.e., structural relationships) and temporal (i.e., historical risk trends) patterns from structured time-series data. The architecture was evaluated using simulated data modeled on real-world settings and its performance was measured using (Root Mean Square Error) RMSE, (Mean Absolute Error) MAE, and (R-Squared) R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> metrics. Results show the superiority of accuracy in the predictions of the hybrid model as compared to traditional ones including (Autoregressive integrated moving average) ARIMA, standalone LSTM, and GNN, thus rendering explicit the model's merits in capturing cascading risks and shifting interdependencies. This study also emphasizes the application of the model in real-time data-driven decision support, with future works looking into deployment on a large scale and real-time external data integration.