Predicting Trends in Tourism Policy Outcomes and Stakeholder Engagement Using Machine Learning in Ghana

Abstract

Effective tourism policy implementation and stakeholder engagement are critical for sustainable tourism development, yet forecasting these dynamics remains a persistent challenge in developing economies such as Ghana, where existing evaluations are largely descriptive. This study applies machine learning to forecast tourism policy outcomes and stakeholder engagement across Ghana’s 16 regions, using 2,500 observations (2016-2025) combining secondary institutional data with primary survey data from policymakers, operators, and community representatives. Four supervised algorithms Random Forest, Gradient Boosting (XGBoost), Support Vector Machines, and Artificial Neural Networks were trained and validated via k-fold cross-validation, with performance assessed using Mean Absolute Error and Root Mean Square Error, and predictors interpreted through explainable feature-importance techniques. Gradient Boosting achieved the highest predictive accuracy. Stakeholder engagement metrics, communication quality, and digital consultation platforms emerged as the strongest predictors of policy outcomes, while funding variables contributed less than expected. Forecasts for 2026-2030 revealed persistent regional disparities, with historically high-engagement regions projected to sustain stronger gains than medium- and low-engagement regions. The study recommended that tourism authorities, such as the Ghana Tourism Authority, prioritise participatory practices and digital engagement platforms alongside financial allocations; that structured stakeholder consultation programmes be institutionalised; and that predictive models be deployed for real-time monitoring to enable adaptive, evidence-based tourism governance in Ghana and comparable contexts.