Abstract for: Forecasting Postal Performance via System Dynamics, Metamodeling and Transfer Learning approach

Postal companies primarily engage in delivering letters, parcels, and providing e-services. As technological advancements continue to emerge and customer behaviors evolve, postal services must adapt and modernize. To gain insights into the postal environment and forecast its performance, we employed system dynamics approach. Our case study focused on Iran Post, examining high-level management at the national level. The simulation model effectively predicts postal service performance. To operationalize this model and leverage high-level management expertise, we applied the Transfer Learning paradigm, tailoring models for 32 distinct regions based on their unique specifications. Additionally, we utilized meta-modeling techniques and Neural Networks to create a mathematical simulation model. Notably, this approach proves particularly effective when there is a large dataset at hand and simulations are very expensive to execute.