DEVELOPMENT A PREDICTIVE OPTIMIZATION MODEL TO MINIMIZE DELAYS AND INEFFICIENCIES AT SPECIAL PORTS

The object of this study is the transportation system of rock aggregates in an archipelagic logistics context, focusing on the integration of sea and land transportation modes at Terminals for Own Use (TOU) in Indonesia. In many island-based countries like Indonesia, maritime logistics efficiency plays a critical role in supporting economic competitiveness. However, challenges such as limited infrastructure, high logistics costs, and low accessibility persist, especially at Terminals for Own Use, where the loading and unloading of bulk commodities like rock aggregates can take 2–5 days due to equipment limitations. These inefficiencies increase mooring times and operational costs, weakening supply chain performance and industrial competitiveness. As demand for construction materials grows, optimizing port infrastructure and transportation connectivity becomes urgent. This study utilizes a Long Short-Term Memory (LSTM) model optimized with Particle Swarm Optimization (PSO) to improve the accuracy of rock aggregate demand forecasting. The model achieves a Mean Absolute Percentage Error (MAPE) of 0.46 % on training data and 5.26 % on test data, indicating high forecast reliability. Time series analysis identifies a downward trend in demand in 2022, indicating the importance of accurate forecasting in reducing inefficiencies. Better forecasting enables better port scheduling and inventory management, leading to a responsive logistics system. The results show that an efficient and demand-responsive transportation system significantly reduces loading time and overall logistics costs. The study highlights that a well-integrated forecasting approach can support better decision-making in port management and transportation planning.