Wang Ruoyao, Shang Wei, Zhang Guoliang
School of Civil Engineering, Architecture and Environment, Hubei University of Technology, Wuhan, 430068, China.
Key Laboratory of Intelligent Health Perception and Ecological Restoration of Rivers and Lakes, Ministry of Education, Hubei University of Technology, Wuhan, 430068, China.
Sci Rep. 2025 Mar 26;15(1):10443. doi: 10.1038/s41598-025-94712-9.
Spontaneous commercial spaces play a crucial role in shaping the vitality of historic districts, yet their spatial characteristics and impact on commercial activity remain understudied. This study employs Mask R-CNN deep learning, random forest regression analysis, and SHAP (Shapley Additive Explanations) to systematically identify and quantify the influence of spontaneous commercial spaces on commercial vitality. Based on a dataset comprising 4217 annotated images collected from Wuhan's Tanhualin Historic District, the study classifies spontaneous commercial spaces into five spatial types and examines their correlation with commercial vitality distribution. The results reveal that convex and scatter-occupying spontaneous spaces have the most significant positive impact, increasing commercial vitality by an average of 22.4% and 16.8%, respectively. SHAP analysis further highlights nonlinear interactions between crowd density and spatial typology, demonstrating that high-density areas amplify the contribution of convex spaces to vitality. Additionally, the spatial density of spontaneous commercial spaces shows a strong correlation with pedestrian flow intensity (R = 0.9062, p < 0.01), indicating their critical role in local economic dynamics. Compared to traditional manual research and spatial analysis methods, the computer vision and interpretable machine learning approaches employed in this study enhance analytical efficiency and causal clarity, providing urban planners with a robust framework for monitoring and evaluating spontaneous commercial spaces. Furthermore, we propose a predictive framework to evaluate the potential of existing commercial streets for future development. The model suggests that areas with 12-18 spontaneous commercial points per 100 square meters exhibit the highest commercial vitality, offering a reference for urban renewal strategies.
自发形成的商业空间在塑造历史街区的活力方面发挥着关键作用，然而其空间特征及其对商业活动的影响仍未得到充分研究。本研究采用Mask R-CNN深度学习、随机森林回归分析和SHAP（Shapley加法解释）方法，系统地识别和量化自发商业空间对商业活力的影响。基于从武汉昙华林历史街区收集的4217张带注释图像的数据集，该研究将自发商业空间分为五种空间类型，并考察它们与商业活力分布的相关性。结果表明，凸形和占据分散区域的自发空间具有最显著的积极影响，分别使商业活力平均提高22.4%和16.8%。SHAP分析进一步突出了人群密度与空间类型之间的非线性相互作用，表明高密度区域放大了凸形空间对活力的贡献。此外，自发商业空间的空间密度与行人流量强度显示出很强的相关性（R = 0.9062，p < 0.01），表明它们在地方经济动态中的关键作用。与传统的人工研究和空间分析方法相比，本研究采用的计算机视觉和可解释机器学习方法提高了分析效率和因果清晰度，为城市规划者提供了一个用于监测和评估自发商业空间的强大框架。此外，我们提出了一个预测框架，以评估现有商业街道未来发展的潜力。该模型表明，每100平方米有12 - 18个自发商业点的区域具有最高的商业活力，为城市更新策略提供了参考。
