Artificial Intelligence–Enabled Modeling of Childcare Service Demand and Policy Efficiency in Inner Mongolia Autonomous Region, China: A Comprehensive Multi-Method Demographic Analysis

Abstract

Demographic contraction is a process that is rapidly unfolding in Inner Mongolia Autonomous Region (IMAR), and is characterized by declining fertility, increasing mortality, negative natural population growth, and a decreasing childcare-age population. These trends present significant challenges for planning childcare, early childhood education and regional fertility policy. This study proposes a forecasting and policy evaluation model based on artificial intelligence and an official panel data set of 2021–2025 with 36 indicators from the Inner Mongolia Statistical Yearbook, China Statistical Yearbook and National Bureau of Statistics communiqués. The framework combines statistical trend analysis, 6 machine-learning regressors, SHAP explainability, BiLSTM forecasting with Monte Carlo dropout, policy-efficiency indices and 5 policy scenarios through 2030. The birth rate for IMAR decreased from 6.26‰ (2021) to 4.28‰ (2025), whereas the natural population growth rate went from 3.76‰ (2021) to −4.58‰ (2025). The number of births used as a proxy for annual childcare demand shrank from 15.0 wan in 2023 to 10.2 wan in 2025, while the population aged 0-2 decreased from 40.4 wan in 2023 to 35.4 wan in 2025. The BiLSTM model further predicts a decrease in births to 9.02 wan by 2028. The machine learning model with the highest performance was Gradient Boosting (RMSE = 1.557 wan; MAPE = 11.33%), which considered birth rate, natural growth rate, and the intensity of childcare as the most important demand predictors, as identified by the SHAP values. A combination of lower policy efficiency and weaker policy-to-outcome conversion has led to the lowest scores for Policy Efficiency Score and Fertility-Friendly Development Index in 2025. A comprehensive intervention that combines a childcare subsidy, an improvement in parental leave, housing policy and protection for women in the labor market can make a difference and bring the projected birth rate to 5.03‰ and the projected demand for childcare to 10.654 wan by 2030, according to scenario analysis. The results offer a basis for adaptive childcare infrastructure, optimising fertility policies, and demographic governance in quickly ageing and low-fertility areas.