Daily forecasting of tourism demand: An ST-LSTM model with social network service co-occurrence similarity

Abstract

In the digital era, social network service (SNS) significantly influences travel behavior. Understanding SNS spillover effects is crucial for accurate tourism demand prediction. This study introduces SNS co-occurrence similarity (SNS-COS) to capture the spillover effects and demonstrates a comprehensive tourism demand forecasting framework that combines econometric and AI models to handle spatial and temporal data. By classifying attractions into different ranks, it effectively captures varying spillover effects. Empirical verification shows significant improvements in forecasting accuracy, especially for low-ranked attractions, providing valuable insights for enhancing tourism demand forecasting and informed decision-making in the tourism industry.