Research on the spatial layout of the elderly care industry based on explainable machine learning: A case study of Hangzhou(基于可解释性机器学习的养老产业空间布局研究)

∶The elderly care industry is significant for addressing the challenges of population aging. However, existing research has not sufficiently explored the factors influencing the spatial layout of elderly care facilities or how to optimally predict these layouts. This study uses Hangzhou's urban area as a case study to fill this gap. Utilizing data from the Seventh National Population Census and POI data, we compare the prediction accuracy of three machine learning models—decision tree (DT), random forest (RF), and extreme gradient boosting tree (XGBoost). The Random Forest model, which showed the highest accuracy, was further analyzed using SHAP to deeply understand the influencing factors. The results indicate that∶ (1) SHAP effectively interprets the results of the machine learning algorithms; (2) Most factors, such as population size and government institutions, positively impact the layout of elderly care facilities, whereas other factors, such as financial service facilities, negatively impact their layout; (3) Future construction of elderly care facilities in Hangzhou should prioritize the central districts of Shangcheng and Gongshu as well as their surrounding areas, while development in Fuyang and Lin'an districts could be moderated. These insights provide a strategic basis for optimizing the spatial distribution of elderly care facilities to better meet the needs of the aging population.∶养老产业发展关乎人口老龄化问题，现有的养老服务设施空间布局影响因素及预测优化研究尚不够深入。以杭州市为例，基于第七次人口普查数据及POI数据，对比了决策树（DT）、随机森林（RF）、极端梯度提升树（XGBoost）3种机器学习模型对养老服务设施空间布局的预测精度，选取较优模型进行空间预测分析，并辅以SHAP模型进行深度解释。结果表明∶（1）RF模型的预测精度较高，SHAP模型可以对机器学习算法进行有效解释；（2）人口数量、政府机构对养老服务设施布局有积极作用，金融服务设施对养老服务设施布局有消极作用；（3）未来杭州市养老服务设施建设重点应布局在上城区、拱墅区及其外围区域，放缓富阳区、临安区的养老服务设施建设。