Social circles, groups, lists, etc. are functionalities that allow users of online social network (OSN) platforms to manually organize their social media contacts. However, this facility provided by OSNs has not received appreciation from users due to the tedious nature of the task of organizing the ones that are only contacted periodically. In view of the numerous benefits of this functionality, it may be advantageous to investigate measures that lead to enhancements in its efficacy by allowing for automatic creation of customized groups of users (social circles, groups, lists, etc). The field of study for this purpose, i.e. creating coarse-grained descriptions from data, consists of two families of techniques, community discovery and clustering. These approaches are infeasible for the purpose of automation of social circle creation as they fail on social networks. A reason for this failure could be lack of knowledge of the global structure of the social network or the sparsity that exists in data from social networking websites. As individuals do in real life, OSN clients dependably attempt to broaden their groups of contacts in order to fulfill different social demands. This means that ‘homophily’ would exist among OSN users and prove useful in the task of social circle detection. Based on this intuition, the current inquiry is focused on understanding ‘homophily’ and its role in the process of social circle formation. Extensive experiments are performed on egocentric networks (ego is user, alters are friends) extracted from prominent OSNs like Facebook, Twitter, and Google+. The results of these experiments are used to propose a unified framework: feature extraction for social circles discovery (FESC). FESC detects social circles by jointly modeling ego-net topology and attributes of alters. The performance of FESC is compared with standard benchmark frameworks using metrics like edit distance, modularity, and running time to highlight its efficacy.
Applicability Evaluation of Several Spatial Clustering Methods in Spatiotemporal Data Mining of Floating Car Trajectory
