Non-Linear Design Thresholds and Configurational Pathways for Positive Emotions in Urban Micro-Renewal: An Explainable Machine Learning and fsQCA Study of Pocket Parks in a High-Density Cold-Climate City

Understanding how built environment features shape visitor emotions through non-linear thresholds and configurational pathways is critical for evidence-based urban micro-renewal in high-density cold-climate cities. However, how these features jointly shape emotional outcomes remains unclear. We analyzed 36,478 geolocated Weibo posts and 16 built environment variables across 142 pocket parks in Shenyang using a three-stage framework integrating Spearman correlation, Random Forest with SHapley Additive exPlanations (SHAP), and fuzzy-set qualitative comparative analysis (fsQCA). Three main findings emerge. First, positive emotions exhibited a V-shaped trajectory (2019–2025) with a summer–autumn peak and winter trough, as well as a core–periphery spatial structure. Online attention was highly skewed, with the top 10 parks accounting for 31% of total posts. Second, land value and road network density exhibited inhibitory thresholds of approximately 9092 RMB/m2 and 0.384 km/km2, whereas internal path density (0.003–0.005 m/m2) and landscape shape index (1.8–2.0) showed optimal ranges. Seasonal analysis further revealed significant threshold contraction, making these constraints notably more restrictive in winter. Third, fsQCA identified eight equifinal pathways, which were grouped into four mechanisms and mapped onto four park types. This study offers a replicable hybrid framework and a Contextual Constraint–Internal Compensation mechanism, providing quantifiable design thresholds and type-specific strategies for cold-region urban micro-renewal.