【摘要】 邻里环境与人类活动行为之间存在非线性关系。本文基于大连市8所公立初中学校采集的青少年活动行为问卷调查数据，利用轻量级梯度提升机算法和可解释性机器学习模型，识别邻里物理和感知环境对青少年上学日和休息日身体活动和久坐行为的非线性影响。结果表明：(1)无论是上学日还是休息日，邻里物理环境对青少年身体活动和久坐行为的集体贡献始终高于邻里感知环境；(2)上学日和休息日，影响身体活动和久坐行为的关键邻里环境特征不完全一致；(3)道路密度、公交车站距离、公园距离是影响上学日身体活动的关键因素，而游乐园距离和停车秩序是影响休息日身体活动的关键因素；(4)对于久坐行为，道路密度、游乐园距离以及公园距离在上学日和休息日都具有关键作用，而绿地仅在上学日相对重要；(5)并非所有邻里环境特征对活动行为的影响都是非线性的，道路密度和游乐园距离与身体活动之间不存在非线性关联。更多还原

【Abstract】 The relationship between neighborhood environments and human activity behaviors is nonlinear. This paper utilizes questionnaire data on adolescent activity behaviors collected from eight public middle schools in Dalian, China. Lightweight gradient boosting machines and explainable machine learning models were employed to identify the nonlinear effects of physical and perceived neighborhood environments on adolescent physical activity and sedentary behavior during weekdays and weekends. The findings indicate that:(1) The collective contribution of the physical neighborhood environment to adolescent physical activity and sedentary behaviors remains higher than that of the perceived environment, both on weekdays and weekends.(2) The key neighborhood features affecting these behaviors differ between weekdays and weekends.(3) Road density, distance to bus stations, and distance to parks are critical factors influencing physical activity on weekdays, while the distance to amusement parks and parking order significantly affect weekend activities.(4) For sedentary behaviors, road density, amusement park distance, and park distance play crucial roles both during weekdays and weekends, whereas green spaces are relatively more important on weekdays.(5) Not all neighborhood features have a nonlinear impact on activity behaviors; for instance, the relationships between road density, amusement park distance, and physical activity are not nonlinear.更多还原