Detection of points of interest from crowdsourced tourism data

Availability of the big data on human mobility raised a lot of expectations regarding the possibility to have a more detailed insights into daily and seasonal mobility patterns. However, this is not a trivial task and often noisy positioning data pose a great challenge among researchers and practitioners. In this paper, we tackle the detection of the Points of Interest (PoI) locations from the mobile sensed tourist data gathered in Zeeland (Netherlands) region. We consider different clustering approaches to detect individuals and collective PoI locations and find that OPTICS proved to be the most robust against initial parameters choices and k-means the most sensitive. K-means also seemed not appropriate to use to extract individual places but it indicates promising to extract areas of city which are often visited.Availability of the big data on human mobility raised a lot of expectations regarding the possibility to have a more detailed insights into daily and seasonal mobility patterns. However, this is not a trivial task and often noisy positioning data pose a great challenge among researchers and practitioners. In this paper, we tackle the detection of the Points of Interest (PoI) locations from the mobile sensed tourist data gathered in Zeeland (Netherlands) region. We consider different clustering approaches to detect individuals and collective PoI locations and find that OPTICS proved to be the most robust against initial parameters choices and k-means the most sensitive. K-means also seemed not appropriate to use to extract individual places but it indicates promising to extract areas of city which are often visited.P