Repository of DUTH

Γραφοθεωρητική ανάλυση κοινωνικών δικτύων για τον υπολογισµό αξιολογήσεων και συσχετίσεων (Master thesis)

Καραγιάννης, Ιωάννης/ Karagiannis, Ioannis

In this work, we show how network analysis can be applied to lists of points of interest (POIs) in order to extract important information about the POIs and the relations between them. More precisely, we construct an affiliation network containing user created lists and POIs, we apply the Hyperlink Induced Topic Search (HITS) algorithm to generate user rating and we propose a Normalized Pointwise Mutual Information (NPMI) based measure to estimate the pairwise similarity between POIs. We evaluate our approach on POIs from the cities of Athens, Thessaloniki, and Rhodes. As a data source we use the corresponding user-specified lists of POIs of Foursquare, which are by default publicly accessible. Our results show that for each POI the authority score obtained with the HITS algorithm is strongly correlated with the actual rating of Foursquare. Moreover, preliminary evidence shows that the NPMI-based measure gives valuable information about the pairwise similarity between POIs.