Abstract: Objective As a vital part of urban green spaces, urban parks provide essential venues for older adults, promoting their physical and mental well-being. Compared to other age groups, older adults individuals have a greater need for engaging in activities within green spaces, and their perception of these environments significantly influences their activity behaviors. Understanding the nonlinear relationship between older adults'' perception of urban park green spaces and their activity characteristics is essential for guiding age-friendly park renovations and effective urban green space planning. Previous studies have examined the impact of green space environments on older adults'' activities, but they often focused on objective environmental features and overlooked subjective perceptions. To address this gap, the present study employs the Extreme Gradient Boosting (XGBoost) model to explore the complex relationship between older adults'' perception of urban park green spaces and their activity behaviors. The study aims to construct a theoretical model that captures the nonlinear dynamics between these variables, using data from urban parks in Fuzhou City for validation. By analyzing the contribution of different perception indicators to older adults'' activity levels, the research seeks to identify threshold effects and provide insights for age-friendly park design and renovation, ultimately supporting the high-quality development of urban green spaces. Methods In this study, older adults'' perception of urban park green spaces was set as the independent variable, while their green space activity levels were defined as the dependent variable. The independent variable, i.e., older adults'' perception of urban parks, includes 12 indicators from the formal questionnaire: accessibility, sense of security, rationality of fitness facilities, rationality of recreational facilities, sanitary conditions, green space quality, maintenance, ecological environment, sound environment comfort, air quality, sense of attachment, and sense of place. These indicators were measured using a 7-point Likert scale. The dependent variable includes four indicators related to green space activity levels: weekly activity frequency, weekly activity duration, activity diversity, and total weekly activity. A total of 779 valid questionnaires were collected from five urban parks in Fuzhou City. The collected data were statistically analyzed using Excel 2013 and SPSS 26.0 to evaluate elderly people''s perception of urban green spaces and their activity levels. In addition, Spearman correlation analysis was conducted to identify the relationship between the independent and dependent variables. For the significant variable group, an interpretable Extreme Gradient Boosting (XGBoost) machine learning model was constructed to explore the relative importance of various perception indicators and the nonlinear relationship between older adults'' perception and their activity characteristics. Results The findings indicate the following: (1) Older adults generally have a positive perception of the green space environment, with air quality and ecological environment receiving the highest evaluations. There were no significant differences in green space perception between different genders and age groups. However, individuals exhibited large differences in their total activity levels, which might be associated with their health conditions, personal interests, and accessibility to green spaces. (2) Sense of place and sense of attachment are the most crucial factors influencing the frequency of older adults'' green space usage, while ecological environment and green space quality significantly impact the duration of use. Additionally, air quality is the most important factor affecting the diversity of activity types, and the ecological environment is the most significant factor influencing the total weekly activity levels. (3) There is a significant nonlinear relationship between older adults'' weekly activity frequency, duration, activity type diversity, and total activity level and several perception characteristics. Among these, sense of place, ecological environment, and green space quality were found to significantly influence activity frequency and duration. Specifically, sense of place had the most pronounced effect on activity levels, while improvements in the ecological environment exhibited an “inverse U-shaped” effect, indicating that continuous enhancement may reduce activity levels after a certain threshold. Moreover, air quality and sound environment were shown to play essential roles in promoting activity frequency and type diversity. Conclusion This study, based on data from elderly populations in five urban parks in Fuzhou, utilizes the XGBoost model to reveal the nonlinear relationship between elderly people''s perception of green spaces and their outdoor activity characteristics. By exploring this relationship from a nonlinear perspective, the study addresses the limitations of traditional research, which often relies on linear analysis models. The findings indicate that factors such as sense of place, ecological environment, air quality, facility rationality, and green space quality are crucial in shaping elderly people''s green space activity patterns. Among these, sense of place has the most significant impact on activity frequency and duration. Additionally, improvements in the ecological environment and green space quality significantly promote outdoor activity participation, although continuous optimization of the ecological environment may lead to a reduction in activity levels, exhibiting an "inverted U-shaped" relationship. Moreover, air quality and acoustic environment further confirm the importance of environmental comfort in enhancing activity frequency and diversity. The rational arrangement and maintenance of fitness and recreational facilities also have multiple impacts on activity duration and variety, with both excessive and insufficient facilities potentially suppressing elderly participation. The study identifies the varying weights and marginal effects of these factors on elderly people''s green space activities, offering a new theoretical perspective for understanding the driving mechanisms behind green space usage. Additionally, the study expands the understanding of the relationship between environmental comfort and activity behavior, providing a new analytical framework and direction for future research. Although this study elucidates the nonlinear relationship between elderly people''s perception of green spaces and their activities, offering scientific evidence for green space usage among the elderly, it is limited by the cross-sectional nature of the data and its regional scope. Future research should incorporate longitudinal designs and cross-regional comparisons to explore the dynamic patterns of elderly people''s green space usage behavior and validate and extend the conclusions of this study. Overall, this research not only clarifies the complex relationship between elderly people''s perception of green spaces and their activity characteristics but also provides scientific support for understanding and optimizing age-friendly urban green space design. Furthermore, it offers policy and practical references for the age-friendly renovation of urban parks, contributing to the high-quality development of green spaces and ultimately promoting elderly health, well-being, and sustainable social development.