Machine Learning Approaches with Random Forest and XGBoost for Sustainable Tourism Forecasting in Bali Destinations
TL;DR
Comparisons of Random Forest and XGBoost algorithms in predicting the popularity of tourist destinations in Bali indicate that nature-based and cultural tourism destinations remain dominant, particularly in Badung, Jembrana, and Gianyar Regencies.
Abstract
Bali has experienced rapid tourism growth, reaching more than 6.3 million international visitors in 2024, which has increased the risk of overtourism and created challenges for sustainable destination management. Despite tourism being a major contributor to Bali’s economy, planning practices have not fully adopted data-driven prediction approaches, resulting in uncertainty in infrastructure development, service capacity, and resource allocation. This study aims to compare the performance of Random Forest and XGBoost algorithms in predicting the popularity of tourist destinations in Bali to support evidence-based decision-making. The research utilizes historical tourist visitation data from 2018 to 2023, obtained from the Bali Provincial Tourism Office. Data preprocessing includes data cleaning, normalization, feature encoding, and dimensionality reduction using Principal Component Analysis. Three data split schemes (80:20, 75:25, and 90:10) are evaluated. Model performance is assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results show that XGBoost outperforms Random Forest, achieving the best performance with a MAPE of 5.32% using the 90:10 data split. The selected model is then applied to project tourism demand for 2024–2026, indicating that nature-based and cultural tourism destinations remain dominant, particularly in Badung, Jembrana, and Gianyar Regencies. This study contributes to informatics by providing a machine-learning-based prediction model to support sustainable tourism management.