The Clusterization of Geo-Tagged Data for Finding City Sights with Use of a Modification of k-MXT Algorithm

Abstract

The paper considers the task of detection of the most attractive for tourists city sights using the database with geotagged data of photographs.We form a graph on the basis of the geo-tagged spot coordinates and rewrite the problem as the graph clusterization task. In our work we use two clustering algorithms, DBSCAN and k-MXT. Moreover, we develop a modification of the k-MXT algorithm called the k-MXT-Gauss algorithm, where the calculation of the weights of the graph edges is transformed using the Gaussian distribution density. We compare the performance of k-MXT-Gauss algorithm with the performance of k-MXT and DBSCAN algorithms both on simulated data and real data.