Mining geosocial data from Flickr to explore tourism patterns: The case study of Athens

Abstract: Kanari Domna Mining geosocial data from Flickr to explore tourism patterns: The case study of Athens Social media are providing a new type of geo-tagged data that by processing them, new types of knowledge can be generated and used by decision-makers in different areas including tourism. It includes e.g., identifying touristic areas that are not prioritized or advertised by touristic authorities. This study aims to investigate a methodology to detect Areas of Interest (AOIs) and temporal distributions of visitors geotagged photos in Athens, by mining and analyzing geosocial data from the Flickr social media platform, from 2009 to 2019. The methodology of this research, divided into 5 stages of procedures: the geosocial data mining, the cleaning process of the data, the spatial clustering analysis, the construction of the database, and the visualization of the results through a Web-GIS application. The total amount of geosocial data harvested from the Flickr social media platform for this research was 157,314 and after the cleaning process was 77,659. To identify the most desired AOIs and the temporal distribution tendencies of the visitors, the HDBSCAN clustering algorithm was applied to the dataset. The algorithm produced 20 spatial clusters in popular areas of Athens and the results of the clustering analysis were stored in a database. To validate the results, 21 of the most famous Points of Interest (POIs) of Athens were gathered, mapped and the correlation between them and the produced AOIs was explored. Finally, the produced AOIs and the collected POIs were presented through a prototype Web-GIS platform among with temporal distribution statistics for each AOI. The results of this study showed that the HDBSCAN algorithm produced 8 new AOIs that were not suggested or advertised by tourism authorities. Also, the findings of this research demonstrate that the study area presents in general medium to high levels of seasonality with small exceptions and that the visitors are mostly from Europe, North America, and Asia. Despite some reliability issues geosocial data present, tourism agencies/authorities, urban planners, and policymakers must seriously consider exploiting such kind of data and understand the power of the tools they can create using location intelligence. Keywords: Geography, Geographical Information Systems, GIS, Spatial Analysis, Spatial Clustering, Density-Based Clustering, HDBSCAN, Web-GIS, Tourism footprints, Social Media data, Geosocial data Advisor: Ali Mansourian Master degree project 30 credits in Geographical Information Sciences, 2021 Department of Physical Geography and Ecosystem Science, Lund University Thesis nr 134