Location-Based Distance Measures for Geosocial Similarity

Abstract

This article investigates the problem of geosocial similarity among users of online social networks, based on the locations of their activities (e.g., posting messages or photographs). Finding pairs of geosocially similar users or detecting that two sets of locations (of activities) belong to the same user has important applications in privacy protection, recommendation systems, urban planning, and public health, among others. It is explained and shown empirically that common distance measures between sets of locations are inadequate for determining geosocial similarity. Two novel distance measures between sets of locations are introduced. One is the mutually nearest distance that is based on computing a matching between two sets. The second measure uses a quad-tree index. It is highly scalable but incurs the overhead of creating and maintaining the index. Algorithms with optimization techniques are developed for computing the two distance measures and also for finding the k -most-similar users of a given one. Extensive experiments, using geotagged messages from Twitter, show that the new distance measures are both more accurate and more efficient than existing ones.

SLIM: Scalable Linkage of Mobility Data

“…Defining a location based similarity between entities is a fundamental problem [17,45]. Some express this based on densities of location histories [13], either by matching user histograms [41], or using the frequencies of visits to specific locations during specific times [35].…”

Efficient Pairwise Penetrating-rank Similarity Retrieval

“…• Adaptivity. Similar to other link-based similarity measures, P-Rank can be combined with other domain-specific similarity measures (e.g., [18]) to produce an overall measure, which is adaptive to any domains with entity-to-entity relationships. • Generality.…”

Infrastructure for discovering geographically related linked data

“…Geo-social similarity refers to the problem of finding similar social network users and some works approach the problem by measuring the distance of point-sets, where a point-set consists of the locations of a user activities. Kanza et al, 2017 propose and evaluate two novel distance measures for finding the k-most similar users to a given user, based on the similarity of their visited places: the mutually nearest distance (MND) and the QuadTree distance (QTD). MND calculates the average distance over the pairs of mutually nearest neighbors, i.e., points, of two point-sets.…”

“…Secondly, they do not take into account the inner-distribution of datasets and most of them are sensitive to outliers. Finally, they have not been evaluated for their effectiveness in the dataset search problem, where a point-set is a collection of the locations of the real-world objects contained in a dataset, and have been proved to be ineffective in some related contexts such as geo-social similarity (Kanza et al, 2017). To tackle the efficiency problem some works (Kanza et al, 2017;Efstathiades et al, 2016), which apply point-set similarity metrics to identify similar users based on their visited locations, propose the use of point-set summaries and indexes.…”

“…The above are used to enable web-scale search in order to identify datasets or classes that contain semantically related real-world objects judging by the similarity of their spatial distribution. Our proposed approach can be useful in other problems that employ point-set similarity, such as geo-social similarity (Kanza et al, 2017), dataset search (Chapman et al, 2019), web tables extension (Lehmberg et al, 2015;Das Sarma et al, 2012), or data source selection for federated queries (Saleem et al, 2014;Vidal et al, 2016;Harth et al, 2010).…”