Spatial Optimization of Primary School Campuses from the Perspective of Children’s Emotional Behavior: A Deep Learning and Machine Learning Approach

From the perspective of children’s emotional behavior, this study constructs a multidimensional indicator framework—“spatial elements-spatial typologies-spatial color-emotion and behavior.” Integrating behavior mapping, we employ deep- and machine-learning models to quantify the pathways through which primary-school campus spaces shape children’s emotional and behavioral responses. The results indicate that: (1) individual external spatial elements exert a more pronounced influence on children’s emotions; (2) different spatial typologies show marked disparities in emotional activation, characterized by polarization and clustering at low levels, revealing common shortcomings in current campus construction; (3) the emotional effects of spatial color diverge systematically by gender and age, with differentiation intensifying as age increases; and (4) overlay analyses of behavior maps corroborate associations between external natural spaces, key internal functional zones, and children’s behavioral patterns and affective responses. Building on these findings, the study proposes targeted optimization strategies oriented toward children’s emotional experience and behavioral development, providing data-driven support for the affective design of primary-school campuses.