Prediction and dynamic analysis of the evolution process of Dongguan red cultural relics based on LSTM model

The evolution of red cultural heritage sites is influenced by various factors, including policy changes, tourist flow, and social development. Traditional research primarily relies on GIS analysis and literature review, lacking precise data-driven predictive models, making it difficult to quantify future trends. To address this issue, this paper proposes a time-series forecasting framework based on Long Short-Term Memory (LSTM) networks, using red cultural heritage sites in Dongguan as the research subject. After data preprocessing, feature extraction, and time window segmentation, the LSTM model is trained and compared with traditional time-series methods such as ARIMA and ETS. Experimental results demonstrate that LSTM outperforms traditional methods in long-term prediction accuracy, reducing Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) by 17.3% and 15.8%, respectively, while increasing the R² score by 12.5%. The study identifies tourist growth and policy support as key influencing factors, with some heritage sites expected to experience a visitor increase of over 30% in the next five years, necessitating enhanced management and protection strategies.