Evaluating Perceived Cultural Ecosystem Services in Urban Green Spaces Using Big Data and Machine Learning: Insights from Fragrance Hill Park in Beijing, China

Cultural ecosystem services (CESs) are essential for the sustainable development and management of urban green spaces. However, there remains a gap in leveraging big data and unsupervised machine learning to comprehensively evaluate perceived CESs. This study introduces a hybrid research methodology integrating latent dirichlet allocation (LDA) and importance-performance analysis (IPA) to analyze 20,087 user-generated reviews of Fragrance Hill Park in Beijing from Meituan. The key findings are the following: (1) ten types of CESs were identified, including five related to personal well-being, four to public well-being, and one bridging both categories; (2) the most significant dimensions were "recreational activities", "aesthetic appreciation", "physical well-being", and "mental well-being"; (3) users expressed positive sentiments toward "history and culture", "mental well-being", and "religious engagement", while "social relations" received the most negative feedback; (4) IPA results highlight "recreational activities" and "aesthetic appreciation" as priority areas for improvement. This study provides a scalable, data-driven framework for evaluating CESs in urban green spaces. The insights gained can inform urban green space management and policy decisions to enhance user experiences and promote sustainable urban development.