User-generated photos in hotel demand forecasting

Abstract

User-generated content has become an invaluable resource for researchers in hospitality and tourism, especially regarding sales and demand forecasting. Some scholars have analyzed textual data and sentiment information; however, few studies have addressed roles of user-generated photos in hotel demand prediction. This study fills this void by examining the effectiveness of various photo features (i.e., topics and sentiments) for hotel demand forecasting. Results demonstrate the superiority of photo topic features over sentiment features in enhancing demand prediction. Forecasting accuracy is further improved after integrating a combination of photo topic and sentiment features. Moreover, user-generated photos elevate the accuracy of daily demand forecasting for different hotels. This study contributes to the literature on hotel demand forecasting using Internet multimodal data.