Hybrid neural network-based metaheuristics in designing robust supply chains under pre-disaster: A case study of blood supply chain

Effective disaster relief requires efficient supply chain management, particularly for perishable and non-perishable goods under uncertain conditions. This study aims to address the challenges of disaster supply chains by proposing a hybrid optimization model that minimizes response times and operational costs, while ensuring the rapid and reliable delivery of essential relief items to affected areas. The model combines predictive analytics and robust optimization techniques in a two-phase approach: pre-disaster planning and post-disaster response. Demand for essential goods is predicted using a hybrid Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) model, optimized with Ant Colony Optimization (ACO) to improve accuracy. The ACO-optimized CNN-RNN model achieved a Mean Squared Error (MSE) of 0.028, Root Mean Square Error (RMSE) of 0.167, and a Coefficient of Determination (R²) of 0.93, demonstrating a 20% improvement in MSE and an 11% reduction in RMSE compared to the unoptimized baseline model. The optimization phase employs Aghezzaf’s robust optimization framework to handle uncertainties in supply, demand, and potential disruptions across the supply chain. The proposed mathematical model categorizes goods into perishable and non-perishable items, incorporates real-world constraints such as inventory expiration and transportation delays, and evaluates performance under various disaster scenarios. The model demonstrates superior predictive performance, achieving significant improvements in accuracy and robustness compared to baseline models. Validation was conducted using statistical tests and real-world scenarios, confirming the reliability of the model. Data authenticity was ensured by sourcing from validated databases and employing cross-referencing techniques for consistency checks. Sensitivity analysis further highlighted the model’s adaptability to different disaster conditions, demonstrating resilience and operational efficiency in minimizing response times and costs. Using the blood supply chain as a case study, the proposed model significantly enhances disaster management by providing a flexible and reliable framework for resource allocation and decision-making. By integrating advanced machine learning techniques with robust optimization, the model bridges the gap between theoretical approaches and practical disaster relief applications. This innovation offers decision-makers a strategic tool for pre-crisis planning, dynamic response strategies, and efficient supply chain operations, contributing to improved outcomes in disaster scenarios.