India’s inbound tourism potential has not been fully realised, even though it has rich heritage resources, expanding air networks, and sustained policy attention. To identify the key determinants of foreign tourism demand, this study applies a novel interpretable machine learning framework, XGBoost with SHAP and ALE plots, on a panel dataset of 61 source countries to India between 2002 and 2024. Using a gravity-based tourism demand model, the analysis uncovers nonlinearities and interactions in tourism demand based on origin-country income, India’s hotel capacity, domestic and international aircraft movements, UNESCO heritage sites, mega-events, inflation, and governance indicators. Accumulated local effects (ALEs) and SHAP values were used as interpretable tools. The results show that source-country income and air connectivity are the most influential drivers of arrivals, while heritage sites and hotel rooms display clear saturation and diminishing returns, and governance and inflation exert only mild or nonlinear effects. Mega-events provide small and inconsistent short-run gains without strong persistence. The findings indicate that India’s future tourism gains lie less in further capacity expansion and more in strengthening air connectivity, strategically targeting emerging middle-income markets, and upgrading quality, governance, and price stability to convert existing assets into sustained spatially dispersed arrivals.
Bidyut Kumar Ghosh (2026) studied this question.