(1) Background: Aging is a global phenomenon, and China’s aging is extensive and rapid and already at the middle to upper level worldwide. Promoting social interaction and increasing positive psychological qualities in individuals are key components in helping people adapt to the physical and mental changes of the aging process. Among them, how middle-aged and older adults improve their physical and mental health through physical activity is of great concern. (2) Methods: This study measured the physical activity of 2721 middle-aged and elderly square dance participants across China, and structural equation modeling was applied to explore the relationship between square dance exercise and group cohesion as well as the role of perceived social support and psychological capital. (3) Results: The results showed that (a) square dance exercise positively predicts group cohesion among middle-aged and older adults. (b) Perceived social support and psychological capital mediate the relationship between square dance exercise and group cohesion, and the mediating role consists of three pathways: perceived social support alone, psychological capital alone, and perceived social support-psychological capital chain mediation. (c) The mediating effect of psychological capital alone is greater than the mediating effect of perceived social support alone and the mediating effect of the perceived social support-psychological capital chain. (4) Conclusions: This study provides support for the theory and practice of square dance exercise and intervention guidance for increasing positive psychological qualities and group dynamic levels in middle-aged and older adults.

Neural architecture search (NAS) has emerged as a promising approach for automating deep learning model design. However, its application in sports analytics faces unique challenges due to the complex interplay between biomechanical patterns, physiological adaptations, and coaching expertise. Traditional NAS methods need help to effectively capture the multifaceted nature of athletic performance, often failing to integrate qualitative coaching insights with quantitative measurements. We introduce ChampionNet, a framework incorporating NAS and large language models to enhance accuracy in predicting athletic performance and tailoring training regimens. Our approach offers three primary contributions: integrating hyperdimensional embedding to capture fine-grained biomechanical features and physiological parameters with exceptional detail, a structure-preserving graph encoding leverages to maintain crucial spatiotemporal relationships in athletic movements, and the novel comprehensiveness of the training graph that models forward performance prediction and backward physiological adaptation pathways. Our experiments on various sports demonstrate that ChampionNet outperforms other models by 2.5% in accuracy and over 61.9% in computational cost. Further insights illustrate the framework's performance with complex patterns and multi-modal data, especially for sports with advanced biomechanical needs. These findings support ChampionNet's effectiveness as an integrative athletic performance optimization solution, highlighting the need for automated architecture search tailored to sports.