AI-Based Demand Forecasting and Resource Allocation for Educational Service Systems: Evidence from Multi-Source International Data

This study develops an artificial intelligence-based forecasting and decisionsupport framework for educational service demand from a service science and informatics perspective.Drawing on multi-source public datasets from the World Bank and UNESCO, we construct a country-level panel dataset covering 50 countries from 2012 to 2022 and develop an Educational Service Demand Index (ESDI) that integrates enrollment participation, education expenditure, teacher availability, digital infrastructure, and socioeconomic conditions.Ridge Regression, Random Forest, and XGBoost models are employed to compare linear and nonlinear forecasting performance under a time-based training and testing strategy.The empirical results show that XGBoost achieves the best predictive performance, with lower MAE and RMSE and higher explanatory power than the benchmark models.Feature importance analysis further indicates that enrollment rates, public education expenditure, and digital infrastructure are the main drivers of educational service demand.The findings provide practical implications for capacity planning, teacher allocation, digital infrastructure investment, and equitable education service delivery.This study contributes to service informatics research by linking AI-based forecasting with public educational service planning and resource optimization.