Housing Quality, Community Environment and Neighborhood Scale

Perceived residential environment quality (PREQ) is used to evaluate neighbourhood liveability, yet survey- based instruments are costly and difficult to update at scale. We propose an LLM-enhanced, PREQ-compatible pipeline to measure neighbourhood perceptions from multi-platform Chinese user-generated text. We analyse 2023-2024 texts from three complementary sources: online reviews, social media posts, and citizen complaints. GPT-4o is prompted to filter PREQ-relevant content, extract evaluation object-content pairs, and score sentiment intensity on a five-point ordinal scale from strongly negative to strongly positive. We compare zero-shot and few- shot prompting with supervised fine-tuning, and benchmark all configurations against two baselines: rule-based dictionary matching and BERT. The fine-tuned model performs best (90.0% object-content extraction accuracy; 92.5% sentiment accuracy), outperforming BERT by 66.6 and 4.5 percentage points, respectively. These outputs support a 45-indicator system mapped to the canonical four-aspect, eleven-scale PREQ framework, with data- driven weights combining mention frequency and sentiment intensity. Nationwide results show a negative- leaning sentiment distribution, with the most salient themes concentrated on environmental health, upkeep and care, and the organisation/accessibility of space, alongside platform-specific emphases. Comparing text salience with an urban-renewal literature corpus reveals attention gaps, especially for operational and man - agement issues. A Beijing case study (2038 communities within the Fourth Ring Road) yields an average PREQ score of 3.40/5, lower in the historic core, and shows a statistically significant but weak positive association with survey-based PREQ, alongside weak correspondence with an objective residential environment index. The framework offers a scalable, resident-centred complement to surveys and objective metrics for neighbourhood diagnostics and renewal prioritisation.

Traditional objective methods for assessing gated community (GC) environments rely primarily on on-site audits and street-view-based virtual audits. The former is costly and labour-intensive, while the latter faces limited spatial coverage and cannot capture temporally variable features. These constraints are especially problematic in enclosed residential forms like GCs, highlighting the need for a more scalable, data-driven assessment approach. This study proposes a multi-source and multi-perspective imagery framework that integrates aerial, external, and internal views to assess GC physical appearance. The framework is structured around a uni fied analysis model with tailored indicators, scoring rules, and integration criteria, enabling standardized measurement across perspectives. This work is methodology-based and is demonstrated through a gated-community case study in Xi ’an District, Mudanjiang City, China. The case study shows the method ’s high agreement with on-site audits while substantially reducing time and labour requirements. Results show that although GCs perform well in security and privacy, they frequently lack adequate public resources and infrastructure. These findings underscore the need for systematic spatial monitoring to identify priority areas for improvement. The ability to incorporate both open-access and actively collected imagery enhances the framework ’s adaptability across diverse urban contexts. This approach provides a standardized, scalable tool for supporting data-driven decision-making in urban planning and residential management. Future extensions may include integrating socioeconomic indicators and applying the method in varied regulatory and geographic settings to further enhance its applicability. 1School of Architecture, Tsinghua University, Beijing, China 2School of Architecture and Hang Lung Center for Real Estate, Key Laboratory of Eco Planning & Green Building, Ministry of Education, Tsinghua University, Beijing, China Corresponding author: Ying Long, School of Architecture and Hang Lung Center for Real Estate, Key Laboratory of Eco Planning & Green Building, Ministry of Education, Room 501, New Architecture Building, School of Architecture, Tsinghua University, Beijing 100084, China. Email: ylong@tsinghua.edu.cn Data Availability Statement included at the end of the article.

校园末端物流无人化的需求测度与模拟优化研究 (申请清华大学工学硕士学位论文 ) 培养单位 ： 建筑学院学科 ： 城乡规划学研究生 ： 梁佳宁指导教师 ： 龙瀛副教授二○二四年五月校园末端物流无人化的需求测度与模拟优化研究梁佳宁 Research on Demand Measurement and Simulation Optimization of Unmanned Last Mile Delivery in Campus Thesis submitted to Tsinghua University in partial fulfillment of the requirement for the degree of Master of Science in Urban and Rural Planning by Liang Jianing Thesis Supervisor : Associate Professor Long Ying May, 2024 学位论文公开评阅人和答辩委员会名单公开评阅人名单武廷海教授清华大学杨滔副教授清华大学答辩委员会名单主席谭纵波教授清华大学委员武廷海教授清华大学田莉教授清华大学龙瀛副教授清华大学郭璐副研究员清华大学张亚轩高级工程师清华大学建筑设计研究院城乡发展规划分院王健教授级高级工程师北京清华同衡规划设计研究院有限公司贾克敬教授级高级工程师中国国土勘测规划院秘书钱辰助理研究员清华大学关于学位论文使用授权的说明本人完全了解清华大学有关保留、使用学位论文的规定，即： 清华大学拥有在著作权法规定范围内学位论文的使用权，其中包括：（1）已获学位的研究生必须按学校规定提交学位论文，学校可以采用影印、缩印或其他复制手段保存研究生上交的学位论文；（2）为教学和科研目的， 学校可以将公开的学位论文作为资料在图书馆、 资料室等场所供校内师生阅读，或在校园网上供校内师生浏览部分内容 ； （3） 按照上级教育主管部门督导、 抽查等要求， 报送相应的学位论文。 本人保证遵守上述规定。 作者签名： 导师签名： 日期： 2024.5.25 日期： 2024.5.25

Background With rapid urbanization, the urban environment, especially the neighborhood environment, has received increasing global attention. However, a comprehensive overview of the association between neighborhood risk factors and human health remains unclear due to the large number of neighborhood risk factor–human health outcome pairs. Method On the basis of a whole year of panel discussions, we first obtained a list of 5 neighborhood domains, con- taining 33 uniformly defined neighborhood risk factors. We only focused on neighborhood infrastructure-related risk factors with the potential for spatial interventions through urban design tools. Subsequently, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a systematic meta-review of 17 infrastructure-related risk factors of the 33 neighborhood risk factors (e.g., green and blue spaces, proximity to major roads, and proximity to landfills) was conducted using four databases, Web of Science, PubMed, OVID, and Cochrane Library, from January 2000 to May 2021, and corresponding evidence for non-communicable diseases (NCDs) was synthesized. The review quality was assessed according to the A MeaSurement Tool to Assess Systematic Reviews (AMSTAR) standard. Results Thirty-three moderate-and high-quality reviews were included in the analysis. Thirteen major NCD outcomes were found to be associated with neighborhood infrastructure-related risk factors. Green and blue spaces or walkabil- ity had protective effects on human health. In contrast, proximity to major roads, industry, and landfills posed serious threats to human health. Inconsistent results were obtained for four neighborhood risk factors: facilities for physical and leisure activities, accessibility to infrastructure providing unhealthy food, proximity to industry, and proximity to major roads. Conclusions This meta-review presents a comprehensive overview of the effects of neighborhood infrastructure- related risk factors on NCDs. Findings on the risk factors with strong evidence can help improve healthy city guide- lines and promote urban sustainability. In addition, the unknown or uncertain association between many neighbor- hood risk factors and certain types of NCDs requires further research. †Yuyang Zhang, Ningrui Liu and Yan Li are joint first authors. *Correspondence: Ying Long ylong@tsinghua.edu.cn Full list of author information is available at the end of the article Page 2 of 15Zhang et al. Environmental Health (2023) 22:2

】研究城市居民居住选择行为的特征规律与发展趋势ꎬ 并预测不同居民群体的居住选址偏好ꎬ对于引导我国城市规划的合理布局、人本发展ꎬ促进城市与居民生活的可持续性ꎬ具有重要意义ꎮ 然而我国以往相关研究多通过传统统计调查数据ꎬ少有大范围、细粒度的研究ꎮ 通过手机信令等多源数据ꎬ可有效提取北京五环内城区大量物理特征、周边设施和社会特征方面的指标ꎬ 探索其空间分布特征与对应关系ꎬ并进一步回答什么样的小区住着什么居民ꎬ什么居民住在什么样的小区此一科学问题ꎮ 【