Extracting Detour Spots Using Skip-gram Model from Geo-tagged Tweets

Numerous contents have been uploaded to social media sites are useful to analyze the tourist behaviors and recommend other tourist spots. In this paper, we propose a method for extracting detour spots from user movement represented by those contents. Our approach uses a new Skip-gram model to learn movements between a pair of locations quantized by their latitude and the longitude. The generated embedding vectors represent the relationships between the movements from one area to the next. We demonstrated that the embedded vectors generated with our proposed method could extract detour spots for a travel route.