Application and Comparative Study of Time Series Analysis Algorithms in Tourist Passenger Flow Forecasting

This paper addresses the challenge of forecasting nonlinear and non-stationary time series in tourist passenger flow. By introducing the EMD-GRU model, it enhances prediction accuracy and stability, supporting optimized tourism resource management. Using daily overnight tourist data from Shanghai (2016–2022), preprocessing steps include missing value handling, outlier detection, and normalization. EMD decomposes time series into multiple IMFs, capturing different frequency components, which are then modeled by GRU to extract temporal dependencies. The final forecast is obtained by reconstructing all IMF predictions and residuals. Comparative analysis with GRU, LSTM, and EMD-RNN demonstrates EMD-GRU’s superiority, achieving the lowest MSE (0.0256), MAE (0.1215), and MAPE (0.0195). Notably, it excels during peak seasons, accurately capturing fluctuations. The EMD-GRU model effectively overcomes limitations of traditional methods in complex seasonal forecasting, offering high adaptability and precision.