微观环境下老旧社区户外空间居民停留偏好模型研究框架构建方法

A Research Framework for the Residence Preference Model of Outdoor Space Residents in OldCommunities under Micro-Environment

关键词:  风景园林  老旧社区  户外空间  居民行为  停留偏好模型  老龄化

Key words:  landscape architecture  old neighborhood  outdoor space  resident behavior  residence preference model  aging

摘要:

社区户外休闲空间(以下简称“户外空间”)是居民利用碎片化时间活动的主要场所，目前高密度城区老旧社区户外休闲空间匮乏，空间供给与居民需 求不匹配，这一问题在老龄化严峻的老旧社区尤其突出。以广州越秀区老龄化老旧社区的户外空间为例，探讨在客观微环境要素影响下的社区户外空间居民停留 行为特征，收集相关数据以机器学习方法建立居民停留偏好预测模型，并用UNA、Envi-met等模拟微观环境数据验证模型，从而构建微观环境影响下老旧社区 户外空间居民停留偏好模型的研究框架。建立和验证模型的结论表明，与研究社区中居民在户外空间停留的相关微环境要素有温度、相对湿度、风速、路径人流 量；XGboost模型用于建立居民停留偏好模型的效果最好，预测系数R 2=0.75，该模型也适用于相似社会场景与建成环境的老旧社区户外空间居民停留情况预 测，排序预测准确率达66.7%。老旧社区户外空间居民停留偏好模型研究框架的建立，为实现特定空间居民停留行为的模拟预测奠定了理论基础，对社区空间更 新有重要意义：一是通过模型预测社区户外空间的居民潜在停留情况，作为社区更新规划布局的科学决策依据；二是借助后期完善的模型定量调整微观环境要素 以提升户外空间的使用效率。

Abstract:

Community outdoor leisure spaces (hereafter referred to as "outdoor spaces") serve as the primary venues for residents to engage in activities during their fragmented free time. Currently, there is a severe shortage of such outdoor spaces in old communities located in high-density urban areas, leading to a significant mismatch between space supply and residents' actual needs. This issue is particularly prominent in aging communities where the elderly population accounts for a large proportion. Taking the outdoor spaces of aging old communities in Yuexiu District, Guangzhou, as a case study, this research explores the characteristics of residents' stay behaviors in community outdoor spaces under the influence of objective microenvironmental factors. Relevant data were collected to establish a prediction model for residents' stay preferences using machine learning methods. Additionally, microenvironmental data simulated by tools such as UNA (Urban Network Analysis) and Envi-met were employed to verify the model, thereby constructing a comprehensive research framework for the residents' stay preference model in old community outdoor spaces under the impact of microenvironments. The conclusions drawn from the establishment and verification of the model indicate that the microenvironmental factors closely related to residents' stay in the outdoor spaces of the studied communities include temperature, relative humidity, wind speed, and pedestrian flow on paths. Among various machine learning algorithms, the XGBoost model demonstrates the optimal performance in constructing the residents' stay preference model, with a prediction coefficient R2 of 0.75. This model is also applicable to predicting residents' stay in outdoor spaces of other old communities with similar social scenarios and built environments, achieving a ranking prediction accuracy of 66.7%. This study establishes a correlation model between the stay preferences of residents in Guangzhou's old community outdoor spaces during summer and objective microenvironmental factors, and verifies the applicability of this model in communities with similar scenario environments, ultimately forming a complete research framework for the outdoor space residents' stay preference model. The establishment of this framework organically integrates behavioral geography methods with spatial planning and design, providing a solid theoretical foundation for the quantitative prediction of residents' stay behaviors in outdoor spaces and decision-making related to spatial renewal. Furthermore, this research holds significant implications for the renewal of community outdoor spaces. Firstly, by predicting the potential stay of residents in community outdoor spaces through the model, it provides a scientific decision-making basis for the layout of community renewal planning, thereby alleviating the imbalance between supply and demand of outdoor spaces. Secondly, it enables the quantitative adjustment of microenvironmental factors through the improved model in the later stage, thereby enhancing the utilization efficiency and vitality of outdoor spaces. This study can be improved in the following two aspects: firstly, expanding the research scope and increasing the number of measured samples for modeling to minimize errors as much as possible; secondly, incorporating subjective factors into the influence mechanism of outdoor space stay to further improve the stay preference model