Κατηγοριοποίηση χρηστών κοινωνικών δικτύων θέσης βάσει χωρο-κειμενικών αποτυπωμάτων
Classification of users of location-based social networks based on spatio-textual footprints
View/
Keywords
Χωροχρονικά δεδομένα ; Κοινωνικά δίκτυα ; Κατηγοριοποίηση χρηστών ; Τροχιές ; Εξόρυξη τροχιώνAbstract
With the development of technology and the increasingly daily use of devices with internet access, a large amount of spatiotemporal data are being generated. This type of data has been of particular concern to the scientific community in recent years. Motion data records the position of a moving object at any moment and is the subject of studying the behavior of not only animals and humans but also solving problems such as predicting the trajectory of a tornado and solving traffic problems.
In this work, the classification of trajectories coming from Twitter users is attempted, using Machine Learning algorithms. Data pre-processing was based on the MasterMovelets method (Ferrero et al., 2020) which discovers relevant sub-trajectories with different and heterogeneous dimensions and varying lengths. The result of the pre-processing step is a data set which is then used as an input for the classification algorithms. The data used come from tweets in Santorini during the period 2018-2019. The analysis showed that the MasterMovelets method may not fit our data as the classifiers gave worse results than those in the work of Ferrero et al (2020). We attempted also classifying without having performed previously any complicated pre-processing. Our aim was not to compare the results coming from these methods as they are fundamentally different. MasterMovelts is based on trajectory mining, while the experimental method classifies trajectory points.