Demand forecasting framework for dynamic fleet sizing decision: a case study of pickup-and-delivery services in Thailand

Accurate demand forecasting is essential for optimizing fleet sizing in pickup-and-delivery services. This study evaluates a demand forecasting framework for dynamic fleet sizing decisions in Thailand's pickup-and-delivery industry. Various models, including ARIMAX, Exponential Smoothing, Random Forest, Multiple Linear Regression (MLR), Artificial Neural Network (ANN), STL Decomposition, and Holt-Winters models, are compared using performance metrics such as MAD, MAPE, MSE, and RMSE. Historical data is used to forecast demand, providing insights for effective fleet sizing. The results highlight the strengths and weaknesses of each model, offering a practical approach for logistics optimization. This research fills the gap in model evaluation and presents a robust framework for demand forecasting in dynamic operational contexts.