Pocket parks have become a popular intervention for improving urban residents’quality of life. Although previous studies have investigated how park attributes contribute to user satisfaction, most assume that they have a (generalized) linear relationship. The assumption may understate the importance of parking attributes that are nonlinearly associated with park satisfaction and misestimate the effects of other attributes, yielding flawed implications for park planning and design. In this study, we employed linear regression and gradient boosting decision trees sequentially on data from Chengdu, a typical large and developing city aiming to become a Garden City by 2035, to examine the associations between park satisfaction and park attributes, which were measured on a five-point scale. Both models showed that the most important correlate of park satisfaction is entrance location, followed by footpath design, trees and shrubs, safety, flowers, convenience for social activities, and greenness.Moreover, some attributes are nonlinearly related to park satisfaction. For instance,green spaces and flowers greatly contribute to park satisfaction only when they perform well (i.e., their performance scores exceed Scale 3 of the five-point scale).The attribute of trees and shrubs diminishes park satisfaction substantially only when its performance is inferior (i.e., its score is lower than Scale 3). Furthermore, some seemingly unimportant attributes (such as recreational facilities and park maintenance) impose a non-trivial detrimental impact on park satisfaction when park users are extremely dissatisfied with the attributes (i.e., their performance scores are at the lowest level). The nonlinear relationships offer nuanced insights on the design of small public parks in densely developed urban areas.