Urban tourist volume forecasting using internet search trends and deep learning methods

Area of Science:

• Tourism Management
• Data Science
• Artificial Intelligence

Background:

• Accurate forecasting of tourist arrivals is vital for urban destination management and marketing.
• Existing methods may not fully capture the dynamic nature of tourism demand influenced by online trends.

Purpose of the Study:

• To develop and validate a novel deep learning framework for predicting tourist arrivals.
• To integrate online search behavior data into tourism forecasting models.

Main Methods:

• Proposed the Dynamic Tourism Network (DTN) model, combining Disentangled Shape and Time series Normalization (Dish-TS) with Temporal Convolutional Networks (TCN).
• Utilized Baidu Index data, reflecting online search trends, for prediction.
• Empirically validated the model on tourist arrival data for Sanya, China.

Main Results:

• The DTN model demonstrated statistically significant improvements in predictive accuracy compared to conventional deep learning approaches.
• The model achieved superior performance across multiple evaluation metrics for tourist volume estimation.
• The integration of online search data proved effective in enhancing forecasting precision.

Conclusions:

• The DTN model offers a robust foundation for real-time tourism demand forecasting in destination management systems.
• Deep learning combined with online search data presents a powerful approach for predicting tourist arrivals.
• Further validation is needed for other destination types and regions with different dominant search engines.