<![CDATA[Tourism is one of the strategic sectors that significantly contributes to regional economic growth, particularly in West Nusa Tenggara (NTB), Indonesia. Accurate forecasting of tourist arrivals is essential to support tourism planning, destination management, and evidence-based policy making. However, conventional forecasting methods often experience limitations in capturing nonlinear and long-term temporal patterns in tourism time-series data. This study proposes a Long Short-Term Memory (LSTM)-based forecasting model optimized using the Adam Optimizer and Gradient Clipping techniques to improve prediction accuracy and training stability. Monthly tourist arrival data consisting of domestic and international visitors during the period of 2014-2023 were obtained from the Tourism Office of West Nusa Tenggara Province. Data preprocessing was performed using Min-Max Scaling before dividing the dataset into training and testing sets with ratios of 70:30 and 80:20. The proposed model was evaluated using the Root Mean Squared Error (RMSE) metric under two training scenarios of 100 and 200 epochs. Experimental results demonstrate that the best forecasting performance was achieved using a 70:30 training-testing ratio with 200 epochs, resulting in the lowest RMSE value of 66.70. The integration of Adam Optimizer and Gradient Clipping improves model convergence stability while reducing prediction errors. Furthermore, the proposed model effectively captures seasonal patterns and long-term trends in tourist arrivals, making it suitable for supporting smart tourism development and intelligent decision-support systems for tourism management in West Nusa Tenggara.]]>
