Tourism planners rely on accurate demand forecasting. However, despite numerous advancements, crucial methodological issues remain unaddressed. This study aims to further improve the modeling accuracy and advance the artificial intelligence (AI)-based tourism demand forecasting methods. Deep learning models that predict tourism demand are often highly complex and encounter overfitting, which is mainly caused by two underlying problems: (1) access to limited data volumes and (2) additional explanatory variable requirement. To address these issues, we use a decomposition method that achieves high accuracy in short- and long-term AI-based forecasting models. The proposed method effectively decomposes the data and increases accuracy without additional data requirement. In conclusion, this study alleviates the overfitting issue and provides a methodological contribution by proposing a highly accurate deep learning method for AI-based tourism demand modeling.

Dining is an essential tourism component that attracts significant expenditure from tourists. Tourism practitioners need insights into the dining behaviors of tourists to support their strategic planning and decision making. Traditional surveys and questionnaires are time consuming and inefficient in capturing the complex dining behaviors of tourists at a large scale. Thus far, the understanding about the dining preferences and opinions of different tourist groups is limited. This article aims to fill the void by presenting a method that utilizes online restaurant reviews and text processing techniques in analyzing the dining behaviors of tourists. The effectiveness of the proposed method is demonstrated in a case study on international tourists visiting Australia using a large-scale data set of more than 40,000 restaurant reviews made by tourists on 2,265 restaurants. The proposed method can help researchers gain comprehensive insights into the dining preferences of tourists.

Approaches to traditional travel diary construction rely on tourist participation and manual recording; hence, they are not only time-consuming but also limited in the scale and the number of samples. Online social network platforms have been used as alternative data sources for capturing the movements and travel patterns of tourists at a large scale. However, they fail to provide detailed contextual information on tourist activities for further analysis. In this paper, we present a new approach to travel diary construction based on the venue check-in data available in mobile social media with rich information on locations, time, and activities. Our case study focuses on the inbound tourism in Hong Kong using a data set composed of 17,355 check-ins generated by 600 tourists. We demonstrate how the proposed travel diary can provide useful practical implications for applications in location management, transportation management, impact management, and tourist experience promotion among others.

The rapid development of the Internet of Vehicles has substantially boosted the prevalence of vehicular social networks (VSN). However, content security has gradually been a latent threat to the stable operation of VSN. The VSN is a time-varying environment and mixed with various real or fake contents, which brings great challenges to the sustainability of VSN. To establish a sustainable VSN, it is of practical value to possess a strong ability for fake content detection. Related works can be divided into the global semantics-based approaches and the local semantics-based approaches, though both with limitations. Leveraging these two different approaches, this paper proposes a fake content detection model based on the mixed graph neural networks (GNN) for sustainable VSN. It takes GNN as the bottom architecture and integrates both convolution neural networks and recurrent neural networks to capture two aspects of semantics. Such a mixed detection framework is expected to possess a better detection effect. A number of experiments were conducted on two social network datasets for evaluation, and the results indicated that the detection effect can be improved by about 5%-15% compared with baseline methods.