In the context of “dual carbon”, parks have shown an essential role in reducing carbon emissions and increasing sinks in cities and have been assigned the function of carbon balancing in the urban carbon cycle. However, while providing leisure and recreational services, urban parks are also accompanied by many carbon emissions. This study aims to reveal whether urban parks are adequate carbon sinks. By constructing a framework for evaluating the carbon balance of urban parks, we assessed the annual carbon balance of 15 urban parks in Nanjing. We used a machine learning method to clarify the key landscape element indicators that affect the carbon balance of urban parks. The results show that there are differences in the annual carbon balance of different urban parks, and most of the parks are “carbon sources”; the larger the area of the park, the higher the vegetation coverage, the stronger the carbon absorption capacity, and lighting, green waste treatment and building operation and maintenance are the primary sources; the three indicators of green space rate, hard surface area, and building area are closely related to the realization of the carbon balance. The three indicators of green space rate, complicated area, and building area are closely related to realizing carbon balance. This study is of positive significance to the achievement of carbon balance in the city, which provides data support for the renewal and renovation of urban parks and provides a reference for the optimization of operation and maintenance management.