Forecasting tourism revenue and evaluating economic impact in Thailand: a machine learning and regional input-output framework

This study introduces the use of alternative data sources combined with machine learning models to forecast provincial tourism revenue in Thailand. The alternative data sources include satellite imagery obtained from Google Earth Engine and tourism-related geospatial data from OpenStreetMap (OSM), which serve as proxies for traditional tourism statistics by capturing the physical and environmental characteristics of tourism-related activities. These data were integrated into four predictive models: Generalized Least Squares (GLS) as the baseline, and three machine learning models—Neural Network (NN), Random Forest (RF), and Support Vector Regression (SVR). The findings reveal that the RF model achieved the highest predictive accuracy, with the lowest root mean square error (RMSE) of 0.0770 and the strongest out-of-sample performance, attaining an R-squared value of 76.05%. Among the explanatory variables, the number of hotels emerged as the most significant predictor of tourism revenue, followed by the presence of restaurants and guesthouses. Additionally, this framework can be further refined to forecast tourism revenue at the district or sub-district level. It is especially useful for monitoring tourism activities in areas where official tourism statistics are limited or unavailable.Beyond forecasting, this study also investigates the broader economic impact of tourism revenue. In Thailand, tourism revenue is heavily concentrated in five provinces—Bangkok, Chiang Mai, Chonburi, Krabi, and Phuket—which together account for approximately 70% of the national total. This concentration poses a challenge to the government’s objective of promoting more equitable regional development. To assess the economic implications, the study employs a Multi-Regional Input-Output (MRIO) framework to evaluate how tourism revenue in these provinces influences output, value-added, and employment across regions. The results indicate that tourism generates substantial economic benefits, though the magnitude of these effects varies by location. Approximately 50%–70% of the total economic gains remain within the province where the revenue originates. Moreover, significant intra-regional spillover effects are observed, representing an additional 10%–20% of the total impact. Sectorally, tourism-related industries—such as accommodation, trade and transportation, and services—absorb 50%–76% of the total economic gains. Among non-tourism sectors, the industrial sector contributes significantly, accounting for 10%–30%, while other sectors experience relatively limited impacts. These spillover effects suggest that high-revenue provinces have the potential to act as strategic hubs for disseminating tourism-related economic benefits to neighboring low-tourism areas. Strengthening supply chain linkages between these tourism hubs and surrounding provinces is therefore essential to achieving a more balanced distribution of tourism gains and advancing inclusive regional economic development.