内嵌时空信息的照片数据中的游览模式探索

Exploring Travel Patterns in Photos with Spatio-temporal Information

随着人们越来越多地在社交网上分享他们的照片,研究者可以获得更多内嵌时空信息的照片数据对用户行为进行挖掘.本文根据在Flickr上获得的用户在北京地区拍摄的照片,发现在该地拍摄照片的用户的停留时间和旅游模式,在此基础上提出一种针对连续拍摄照片的时间分割准则,并基于密度峰值聚类定义拍摄兴趣区和游览路径(关键径).为了更准确地表达用户游览兴趣区的偏好,还定义了紧邻后向频率,紧邻前向频率,全局后向频率和平均全局后向频率.基于在北京地区拍摄的20万张照片的时空信息,使用提出的方法对游客游览北京的模式进行了分析,验证了该方法的可行性与合理性.

More and more people started to upload more photos, which enables researchers to get more spatio-temporal data for analy- zing users＇ behavior. We fetched photos from Flickr taken in Beijing, and found that most of the authors of these photos didn＇t spent a long time in Beijing. Based on this, this paper proposed a new criterion for segmenting a user＇s photo sequence, and presented a defini- tion of regions of interest and travel path based on DPC clustering algorithm. In order to present tourists＇ travel behavior more precise- ly, we also defined next visit frequency ,previous visit frequency, global next visit frequency and average global next visit frequency. Based on 200 thousand photos＇ spatio-temporal information taken in Beijing, we made an analysis on tourists＇ travel behaviors, and the feasibility and rationality is verified.