OPTIMALISASI DISTRIBUSI MULTI-WAREHOUSE MELALUI INTEGRASI REINFORCEMENT LEARNING DAN BIG DATA BERBASIS SIMULASI MONTE CARLO

Multi-warehouse distribution faces persistent challenges such as stock imbalance, inefficient routing, and demand uncertainty that are difficult to address using conventional methods. This study develops an adaptive optimization model that integrates Reinforcement Learning, Big Data analytics, and Monte Carlo simulation to overcome these limitations. A simulation-based experimental design is employed using synthetic data representing a network of 10 warehouses, 200 customers, and stochastic demand patterns. A Deep Q-Network model is constructed to generate adaptive distribution policies, while Spark Streaming is used to simulate real-time demand updates. Evaluation across 1,000 Monte Carlo scenarios shows that the model maintains high distribution efficiency, improves demand prediction accuracy, and achieves more stable on-time delivery compared to static routing approaches. These findings demonstrate that integrating RL, Big Data, and stochastic simulation enhances system resilience under dynamic operational conditions. Theoretically, the study contributes to logistics and RL research by emphasizing the importance of Big-Data-driven state representation and probabilistic validation. Practically, the model offers potential for adoption by logistics companies seeking to improve cost efficiency, service quality, and operational adaptability. Overall, the study highlights the effectiveness of combining RL, Big Data, and Monte Carlo simulation as a computational approach for optimizing multi-warehouse distribution systems.