As economic growth and societal shifts reshape urban environments, cultural and creative industrial parks are emerging as vital contributors to sustainable urban development. The design of these landscapes plays a pivotal role in enhancing user satisfaction, increasing spatial attractiveness, and promoting eco-friendly urban practices. This study examines visitor landscape perception preferences in the Textile and Garment Cultural and Creative Industrial Park, located in Haizhu District, Guangzhou, through a novel methodology combining user-generated content (UGC), deep learning models, outdoor electrodermal activity (EDA) measurements, and questionnaire surveys. The UGC-based landscape recognition model achieved an accuracy of 86.8% and was validated against user preferences captured through questionnaires. Results demonstrate that visitors prefer areas featuring cultural landmarks and natural elements, while spaces dominated by human activity and transportation infrastructure are less favored. Key landscape elements, such as signage, thematic sculptures, brand logos, and trees, were identified as highly preferred features within the park. While EDA experiments revealed significant variations in physiological responses across different spatial settings, no strong correlation was observed between EDA indicators and subjective questionnaire scores. This integrative approach enables a comprehensive, objective assessment of landscape perception, providing a data-driven, user-centered framework for improving landscape design in cultural and creative industrial parks.