Mining tourists’ destinations and preferences through LSTM-based text classification and spatial clustering using Flickr data

Recently, a large volume of data related social network service (SNS) is being produced as mobile devices are evolved and SNS is being used ubiquitously. People usually refer to social media when choosing tourist destinations and deciding on tourism activities. Flickr data has been widely utilized in the study of tourism. However, existing studies have limitations in covering the characteristics of tourism activities. In this study we initially developed a tourism category with topic modeling, classified Flickr text data with long short term memory (LSTM) according to the tourism category, and then derived region of attractions (ROA) of each tourism category by spatial clustering analysis, and finally identified attractive factors for each ROA. In this study, we derived nine tourism categories and found that the attractive factors for each ROA were different for each of tourism categories. This study is significant in that it is possible to analyze tourists' preferences in detail by combining deep learning-based text classification and spatial data analysis. In addition, framework and findings proposed in this study can be applied in other urban studies as well as tourism management.