Behavior Analysis of Photo-Taking Tourists at Attraction-Level Using Deep Learning and Latent Dirichlet Allocation in Conjunction With Kernel Density Estimation

Abstract

User-generated content (UGC) on social media platforms plays a significant role in conveying individual sentiments that are effective in predicting image sentiments by evaluating its contents to excavate the behaviour and cognition of image producers at attraction-level tourist destinations. In this study, we aim at the attraction-level study of the spatiotemporal behavior of photo-taking tourists. We used the deep convolutional neural network model DeepSentiBank (DSB) for sentiment prediction of Flickr photos. Then, Latent Dirichlet Allocation (LDA) was employed to categorize these sentiment predictions according to the noun content of specified distinct groups. Next, Kernel Density Estimation (KDE) was used as a spatial analysis tool to determine the spatial distribution characteristics of tourists' behaviors. The photo-taking behavioural patterns of different tourist types are analyzed from the individual preferences of tourists and depicted in terms of both visual semantics and spatiotemporal behavioral features. It is observed that the proportion of landscape preferences of the thematic tourists are not significantly affected by seasons but by the aggregate activities of thematic tourists that significantly vary in autumn. The findings obtained from the analysis hold immense significance in the realm of tourism and hospitality management which can play a significant role in the development of facility management, regional tourism, and prospects.