仮想実社会データを用いたリアルスケール社会シミュレーションの実現

Real-Scale Social Simulation (RSSS) often cannot directly use detailed individual microdata from official registers or censuses in large-scale agent-based simulations due to access restrictions and privacy concerns. This paper organizes key policies and enabling techniques for constructing “virtual real-world data,” i.e., synthetic populations coupled with basic behavioral data, from publicly available statistics. Using sample-free simulated annealing, nationwide household and individual attributes (e.g., age, sex, industry) are synthesized to match statistical constraints, together with procedures for allocating agents to buildings and for defining appropriate usage conditions with privacy considerations. In addition, the study links population census data with the Economic Census to estimate home–work locations of employees at small-area resolution, extending static nighttime population distributions to dynamic daytime distributions. The proposed data foundation supports high-fidelity evaluation of policies and services such as travel demand estimation, disaster impact assessment, and telework support during pandemics.