<![CDATA[Tourism demand forecasting remains challenging due to the dynamic nature of consumer behavior and the lag in traditional economic indicators. This study investigates the predictive power of online search behavior captured through Google Trends in forecasting international tourist arrivals to Indonesia, integrating digital footprints with macroeconomic indicators using machine learning approaches. Monthly data spanning January 2011 to December 2024 (N=168) were collected from multiple sources: tourist arrivals from Statistics Indonesia, Google Trends composite index from 11 destination-related keywords, exchange rates, and consumer price index. Three machine learning algorithms Random Forest, XGBoost, and LSTM were compared using temporal train-test split (80:20). Random Forest demonstrated superior performance with Test MAPE of 14.12% and R² of 0.811, outperforming XGBoost (MAPE 21.41%) and LSTM (MAPE 24.56%). Feature importance analysis revealed that Google Trends emerged as the second most important predictor (12.22% contribution), surpassing all economic indicators combined (7.20%), validating its role as a leading indicator of tourism demand. This study contributes methodologically by providing empirical evidence for the value of integrating digital behavioral signals with traditional predictors in emerging destination contexts, and practically by demonstrating deployment-ready forecasting accuracy for tourism stakeholders in Indonesia.]]>
