| 摘要: |
| 滨海空间衔接城市与水域，是滨海
游憩活动的重要载体。构建包含游憩者多维
感知体系，基于社交媒体的多模态感知数
据，采用SHAP可解释性分析方法，分析景
观感知要素对大连市滨海游憩空间用户感知
的非线性影响。研究表明：第一，滨海游憩
空间用户感知呈现显著非线性特征，情感感
知维度预测性最佳，学习认知维度非线性效
应最强；第二，建筑覆盖率为最关键影响因
素，贡献峰值位于10%~15%，灰空间比对学
习认知与消费服务作用突出，自然可视率对
情感感知影响显著；第三，识别出四种空间
类型，分别为消费、认知、氛围、情感-感官
主导型。针对不同类型提出系统性环境质量
提升、多要素平衡配置等差异化规划策略，
为滨海游憩空间视觉优化提供借鉴。 |
| 关键词: 滨海游憩空间 社交媒体数据 景
观要素 感知维度 机器学习 |
| DOI：10.13791/j.cnki.hsfwest.20251012005 |
| 分类号: |
| 基金项目:国家自然科学基金面上项目（52278048） |
|
| A study on the nonlinear impact of visual experience on user perception in coastal recreationspaces: An empirical study in Dalian based on social media data |
|
CAI Jun,XU Jiaqi,WU Yishuang,DING Mengzhen
|
| Abstract: |
| Coastal recreational spaces serve as essential connectors between urban areas and natural
water bodies, providing significant opportunities for various recreational activities. These spaces are
crucial for public well-being, allowing individuals to engage with nature and enjoy outdoor
experiences. As urbanization accelerates and the demand for high-quality public spaces increases,
optimizing the layout and design of coastal recreational areas has become essential. While existing
studies have mainly focused on users’ subjective perceptions through traditional methods such as
surveys and interviews, few have explored the nonlinear relationships between visual landscape
elements and user perceptions. This study bridges this gap by investigating the nonlinear impacts of
visual experience on user perception in coastal recreational spaces, utilizing a novel approach
integrating multimodal social media data and SHAP (Shapley Additive Explanations) analysis.Using
social media data, including text, images, and geotagged check-ins, this research constructs a multidimensional
perception system to explore how various landscape elements influence user experiences
in coastal recreational spaces in Dalian, China. Social media data provides a rich source of real-time
user feedback, enabling the analysis of user perceptions across different spatial and temporal contexts.
The study employs SHAP to quantify the influence of nine key landscape perception factors, offering
a deeper understanding of how each element contributes to users’ perceptions in a nonlinear manner.
This methodology allows the identification of both direct and indirect effects of these factors,
presenting a comprehensive picture of how coastal spaces are perceived by the public. The results
reveal several key findings. Firstly, user perception of coastal recreational spaces exhibits significant
nonlinear characteristics. Emotional experience is the most predictive factor, meaning that users’
emotional responses—such as feelings of relaxation or enjoyment—are critical in shaping their
overall perception. The learning and cognitive dimension shows the strongest nonlinear effects,
suggesting that landscape features like natural visibility significantly enhance users’ cognitive
engagement, but only after certain thresholds are reached. This challenges traditional linear models of
landscape perception, where the relationship between environment and perception is assumed to be
proportional.The study identifies building coverage as the most influential landscape feature affecting
user perception. The marginal contribution of building coverage peaks at around 10%~15%,
indicating that moderate development optimizes user experiences. Exceeding this range leads to a
decline in perceptions, suggesting that overdevelopment may detract from spatial quality. The grey
space ratio—areas not easily accessible or usable—affects learning cognition and consumer services.
Additionally, natural visibility, referring to the extent to which users can see natural elements like
water, trees, and vegetation, significantly impacts emotional perception. The greater the visibility of
natural elements, the more positive the emotional response. Four distinct spatial types are identified
consumption-dominant, cognition-dominant, atmosphere-dominant, and emotional-sensory dominantspaces. Each spatial type is characterized by a different dominant perception dimension. For instance, consumption-dominant spaces are marked by high
satisfaction with commercial services, while cognition-dominant spaces focus on intellectual engagement. Atmosphere-dominant spaces evoke strong ambiance
due to natural elements, and emotional-sensory dominant spaces engage users’ emotions and sensory responses through both architecture and nature. |
| Key words: coastal recreation space social media data landscape perception experience dimensions machine learning |
