AI 中文总结
针对劳动年龄人口下降问题，提出基于QUBO框架探索社会指标配置，利用日本市政数据回归建模，经多种方法优化，该框架能连接多方面因素，提供可解释方法，用于政策讨论探索。
AI 中文摘要
劳动年龄人口下降是区域可持续发展的重大挑战，尤其在日本这样的老龄化社会。我们展示了一个基于二次无约束二进制优化（QUBO）框架的方法，用于探索与劳动年龄人口增长相关的社会指标配置。利用日本市政数据，我们将2010 - 2020年劳动年龄人口增长率对十个离散化社会指标进行回归。所得二次替代模型具有合理预测性能，测试集相关系数为0.84，平均R平方值为0.76。其系数矩阵提供了个体指标贡献和成对关联的可解释表示。我们将拟合模型转换为带独热约束的QUBO公式，并使用量子退火、模拟退火和Gurobi进行优化。三种方法都确定了相同的最优可行配置，基于退火的采样器还生成了具有不同预测增长率的可行次优配置。市级单指标分析表明，改变一个指标会根据其他指标增加或降低预测增长率。该框架为连接市政社会统计、非线性相互作用和基于模型的情景生成提供了一种可解释且易于优化的方法，应被视为政策讨论的探索工具而非政策干预的因果估计。
英文摘要
The decline of the working-age population is a major challenge for regional sustainability, particularly in ageing societies such as Japan. We present a methodological demonstration of a quadratic unconstrained binary optimization (QUBO)-based framework for exploring social-indicator configurations associated with working-age population growth. Using Japanese municipal data, we regressed the 2010-2020 working-age population growth rate on ten discretized social indicators. The resulting quadratic surrogate model showed reasonable predictive performance, with a test-set correlation coefficient of 0.84 and an average R-squared value of 0.76. Its coefficient matrix provides an interpretable representation of individual indicator-level contributions and pairwise associations. We converted the fitted model into a QUBO formulation with one-hot constraints and optimized it using quantum annealing, simulated annealing, and Gurobi. All three methods identified the same optimal feasible configuration, while the annealing-based samplers also generated feasible suboptimal configurations with different predicted growth rates. Municipality-level single-indicator analyses showed that changing one indicator can increase or decrease the predicted growth rate depending on the other indicators. The framework provides an interpretable and optimization-ready approach for connecting municipal social statistics, nonlinear interactions, and model-based scenario generation. It should be regarded as an exploratory tool for policy discussion rather than as a causal estimate of policy interventions.
Comments10 pages, 3 figures
