Supply Chain Optimization and Resilience of Instant Noodles Using a Hybrid System Dynamics (SD) and Long Short-Term Memory (LSTM) Modeling at PT Karunia Alam Segar, Gresik

Abstract

This study aims to develop a hybrid System Dynamics (SD) and Long Short-Term Memory (LSTM) model to enhance demand forecasting accuracy and supply chain resilience for instant noodles at PT Karunia Alam Segar, Gresik. Daily historical data from the 2021–2025 period were processed using Python and Vensim PLE through a bidirectional integration that combines the predictive capability of LSTM with the structural system analysis of SD. The results demonstrate that the hybrid model achieved the highest accuracy, with a Mean Absolute Percentage Error (MAPE) of 4.23% and a Root Mean Square Error (RMSE) of 38.1 tons/day, outperforming the standalone SD model (10.42%; 89.3 tons/day) and the standalone LSTM model (5.27%; 45.6 tons/day). The implementation of this model reduced the risk of raw material shortages to 3.7%, decreased overstocking by 31.8%, yielded annual operational cost savings of 15.2%, and suppressed the Bullwhip Effect by up to 26.4%. Furthermore, the model proved resilient when subjected to demand surges of up to 25% and supply delays of 15 days. This approach is effective for large-scale food manufacturing applications and supports sustainability targets in alignment with SDG 9, SDG 12, and SDG 13.