Enhancing Tourism Demand Forecasting with a Transformer-based Framework

Enhancing Tourism Demand Forecasting with a Transformer-based Framework

2024 | Xin Li, Yechi Xu, Rob Law, Shouyang Wang
This study introduces an innovative framework that leverages the latest transformer architecture to enhance tourism demand forecasting. The proposed model integrates the tree-structured parzen estimator for hyperparameter optimization, a robust time series decomposition approach, and a temporal fusion transformer for multivariate time series prediction. The method first decomposes the data series to mitigate the impact of outliers using a decomposition method, followed by forecasting with the temporal fusion transformer, whose hyperparameters are fine-tuned using a Bayesian-based algorithm. This approach surpasses existing state-of-the-art methodologies in terms of forecasting accuracy and robustness. The study demonstrates the effectiveness of the proposed model on three datasets: Hawaii, Jiuzhaigou, and Mount Siguniang, showing significant improvements in forecasting accuracy and stability compared to other benchmark models. The Diebold–Mariano test further confirms the superior performance of the proposed model. Interpretability analysis through attention weights and variable importance provides insights into the model's decision-making process, highlighting its potential as a valuable tool for tourism researchers and practitioners.This study introduces an innovative framework that leverages the latest transformer architecture to enhance tourism demand forecasting. The proposed model integrates the tree-structured parzen estimator for hyperparameter optimization, a robust time series decomposition approach, and a temporal fusion transformer for multivariate time series prediction. The method first decomposes the data series to mitigate the impact of outliers using a decomposition method, followed by forecasting with the temporal fusion transformer, whose hyperparameters are fine-tuned using a Bayesian-based algorithm. This approach surpasses existing state-of-the-art methodologies in terms of forecasting accuracy and robustness. The study demonstrates the effectiveness of the proposed model on three datasets: Hawaii, Jiuzhaigou, and Mount Siguniang, showing significant improvements in forecasting accuracy and stability compared to other benchmark models. The Diebold–Mariano test further confirms the superior performance of the proposed model. Interpretability analysis through attention weights and variable importance provides insights into the model's decision-making process, highlighting its potential as a valuable tool for tourism researchers and practitioners.
Reach us at info@study.space