Today, logistic systems face many issues. Logistics terminals increasingly suffer from congestion, long waiting times and inefficient truck scheduling. To overcome this limitations this paper proposes a closed loop framework combining Machine Learning (ML)-based Estimated Time of Arrival (ETA) prediction with dynamic dock scheduling and rescheduling at logistics terminals under truck arrival uncertainty. Three prediction models namely an Artificial Neural Network (ANN), a Random Forest (RF), and XGBoost are tested on a dataset from the U.S. Bureau of Transportation Statistics (BTS) adapted to a freight logistics. Among the three tested models, XGBoost achieves the best performance with a Mean Absolute Error (MAE) of 6.35 minutes and an R2 of0.9768,andisselectedas an inputto MixedIntegerLinearProgramming (MILP) formulations models: an initial scheduling model minimizing makespan based on predicted ETAs, and a dynamic rescheduling model that adjusts assignments based on Actual Time of Arrivals (ATAs). Validated on five days from the test set with 6 docks. These Results demonstrate that tightly coupling data-driven ETA prediction with proactive scheduling decisions significantly reduces terminal disruption.