Abstract | ||
---|---|---|
The accumulation of users' whereabouts in location-based applications has made it possible to construct user mobility profiles. Trajectory patterns, i.e., traces of places of interest that a user frequently visits, are among the most popular models of mobility profiles. In this paper, we revisit measuring user similarity using trajectory patterns, which is an important supplement for friend recommendation in on-line social networks. Specifically, we identify and formalise a number of basic principles that should hold when quantifying user similarity with trajectory patterns. These principles allow us to evaluate existing metrics in the literature and demonstrate their insufficiencies. Then we propose for the first time a new metric that respects all the identified principles. The metric is extended to deal with location semantics. Through experiments on a real-life trajectory dataset, we show the effectiveness of our new metrics. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1007/978-3-319-11116-2_38 | WEB TECHNOLOGIES AND APPLICATIONS, APWEB 2014 |
Field | DocType | Volume |
Data mining,Social network,Computer science,Trajectory,Database,Semantics | Conference | 8709 |
ISSN | Citations | PageRank |
0302-9743 | 7 | 0.49 |
References | Authors | |
10 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xihui Chen | 1 | 92 | 7.22 |
Ruipeng Lu | 2 | 7 | 0.49 |
Xiaoxing Ma | 3 | 511 | 57.89 |
Jun Pang | 4 | 219 | 33.53 |