This research paper proposes an AI-based predictive model that uses deep learning and text mining techniques to forecast visitor numbers at art museums. The study employs eight deep learning algorithms—including RNN, LSTM, and Transformer—to analyze unstructured textual data from museum websites, visitor comments, and social media, integrating a Balanced Scorecard framework with four strategic perspectives: social value, visitor experience, exhibition management, and art education.
Why it matters: While AI has been adopted in other cultural sectors, art museums have lagged in using deep learning for attendance prediction. This study fills that gap by demonstrating how qualitative, text-based data can enhance forecasting accuracy beyond traditional numeric models. If implemented, such models could help museums optimize resource allocation, personalize visitor engagement, and align operations with their cultural missions, potentially transforming how institutions plan exhibitions and manage audiences.