Design of Machine Learning Algorithm for Tourism Demand Prediction.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Tourism Management

Background:

• Accurate tourism demand forecasting is crucial for governmental planning, infrastructure development, and lodging site selection.
• Wasted resources from unsold inventory highlight the economic importance of precise demand prediction.
• Advancements in Artificial Intelligence (AI) have led to the effective application of models like neural networks in tourism forecasting.

Purpose of the Study:

• To develop an improved machine learning model for tourism demand forecasting.
• To enhance the performance of deep learning models in predicting tourist arrivals.
• To validate the proposed model's superiority over existing methods.

Main Methods:

• A novel stacked autoencoder integrated with long short-term memory (SAE-LSTM) neural networks was constructed.
• A hierarchical greedy pretraining method was employed for deep network weight initialization, replacing random initialization.
• Monthly tourism volume and related influencing factor data, including search engine trends, were utilized for model training and validation.

Main Results:

• The proposed SAE-LSTM model demonstrated superior prediction performance compared to the standard Long Short-Term Memory (LSTM) model.
• Comparative experiments using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and MAPE confirmed the enhanced accuracy of the SAE-LSTM model.
• The unsupervised pretraining method based on LSTM proved effective in improving deep learning model performance for tourism forecasting.

Conclusions:

• The developed SAE-LSTM model offers a more accurate approach to tourism demand forecasting.
• The unsupervised pretraining strategy significantly enhances the predictive capabilities of deep learning models in this domain.
• This research provides a valuable tool for governments and tourism stakeholders to optimize planning and resource allocation.