How does the perception of informal green spaces in urban villages influence residents’ complaint Sentiments? a Machine learning analysis of Fuzhou City, China

• Published in: Ecological indicators 2024-09, Vol.166, p.112376, Article 112376
• Main Authors: Chen, Zhengyan, Yang, Honghui, Ye, Peijin, Zhuang, Xiaowen, Zhang, Ruolan, Xie, Yuanqin, Ding, Zheng
• Format: Article
• Language: English
• Subjects: Informal Green Spaces; Machine learning; Residents’ Complaint; Sentiment analysis; Urban Villages
• ISSN: 1470-160X
• DOI: 10.1016/j.ecolind.2024.112376

•More resident environmental complaints are concentrated near the city center.•Greenness and paving degree have the most significant impact on complaint emotions.•Enclosure, perception sentiment, and color complexity have a noticeable positive impact.•Greenness shows instability during indicator interactions. Informal Green Spaces (IGS) are unplanned and often overlooked vital green resources in urban environments. Previous research suggests that IGS can affect residents’ emotions and generate complaints. However, further research is needed to determine the specific impact of detailed indicators within the IGS environment on changes in residents’ complaint emotions. This study employs machine learning models such as Long Short Term Memory networks-Convolutional Neural Networks (LSTM-CNN), Object Semantic Attention Network (OSANet), High-Resolution Net (Hrnet), and EXtreme Gradient Boosting (XGBoost), combined with hotspot analysis and SHapley Additive exPlanation (SHAP) methods. We revealed the potential impact of urban village IGS landscapes on residents’ complaint sentiments. The results indicate: (1) Complaint sentiment text analysis shows that negative emotions in residents’ complaints account for 85.4%, with hotspot areas spreading outward from the city center. The spatial autocorrelation of IGS indicators shows a strong clustering effect, with significant changes at the boundaries of hotspot areas near the city center and in the northeast. (2) Greenness (1.19), Paving degree (0.93), Openness (0.86), and Color complexity (0.84) emerge as the four most impactful indicators on the complaint sentiments of urban village residents. Enclosure, Perception sentiment, Color complexity, and Paving degree significantly contribute to the model. (3) Greenness has the most substantial impact on emotional changes when interacting with other landscape elements, a higher Color complexity value may lead to negative effects, while Perception sentiment, Enclosure, and Greenness exhibit neutralizing effects on emotions when combined with other indicators. This study proposes a framework for IGS data acquisition and assessment, integrating the strengths of different machine learning methods. By doing so, it provides a data foundation for the optimization and renewal of IGS in urban villages, fully exploring the potential of urban village IGS in urban renewal.