LLMとPOIデータを用いた東京における都市機能地域の同定
実施中
関本 義秀
Accurate identification of Urban Functional Areas (UFAs) is essential for evidence-based urban planning and the promotion of sustainable development. This study introduces a pioneering framework for Tokyo that synergistically integrates graph-based structural learning with the semantic intelligence of Large Language Models (LLMs). While traditional methods often rely on simple point-of-interest (POI) counts, this approach captures complex spatial dependencies and the nuanced linguistic context of urban activities, leading to a more sophisticated understanding of how city spaces are actually utilized. Beyond technical accuracy, the research focuses on enhancing urban livability and social equity by meticulously analyzing service accessibility and the distribution of essential amenities. By identifying intricate mixed-use zones and evolving functional patterns, this work provides policymakers with critical insights needed to optimize infrastructure investments and transportation design. These insights are particularly vital for supporting contemporary urban concepts like the "15-minute city," which aims to ensure that residents have equitable access to everyday needs—such as healthcare, education, and grocery stores—within a short walk or cycle from their homes. Ultimately, the integration of semantic and spatial data facilitates more informed and socially impactful urban interventions. By bridging the gap between high-dimensional data analysis and human-centric planning, this framework offers a scalable solution for creating more resilient, inclusive, and efficient metropolitan environments in the face of rapid urbanization.
変更のために新しい申請を保存します。 This will save a new application on the system for a modification.
申請中の研究者は表示されません。 / Pending researchers are not shown.
鐘 志華 / 東京大学 空間情報研究センター 関本研究室
申請中のデータセットは表示されません。 / Pending datasets are not shown.
座標付き電話帳DBテレポイント 法人版（P1B08_2021年8月）
年次報告の内容はメンバーのみ表示されます。