ABSTRACT
The great number of social network users and the expansion of this kind of tool in the last years demand the storage of a great volume of information regarding user behaviour. In this article, we utilise interaction records from Facebook users and metrics from complex networks study, to identify different user behaviours using clustering techniques. We found three different user profiles regarding interactions performed in the social network: viewer, participant and content producer. Moreover, the groups we found were characterised by the C4.5 decision-tree algorithm. The 'viewer' mainly observes what happens in the network. The ‘participant’ interacts more often with the content, getting a higher value of closeness centrality. Therefore, users with a participant profile are responsible, for example, for the faster transmission of information in the virtual environment, a crucial function for the Facebook social network. We noted too that ‘content producer’ users had a greater quantity of publications in their pages, leading to a superior degree of input interactions than the other two profiles. Finally, we also verify that the profiles are not mutually exclusive, that is, the user of a profile can at determined moment perform the behaviour of another profile.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Pedro Henrique B. Ruas http://orcid.org/0000-0001-6423-8681
Notes
1 Data from Alexa.com. Accessed on April 7, 2017.
2 The BetaCV is the ratio between the average intra-cluster distance and the average inter-cluster distance. The smaller the BetaCV value, the better the cluster, since it indicates that the intra-cluster distances are smaller than the inter-cluster distances (Zaki and Meira Citation2014).
3 A location-based online social network based which combines SNSs with geographic information sharing. Available at: https://pt.foursquare.com/.
4 Available in: https://developers.facebook.com/docs/graph-api
5 Available in: https://nodexl.codeplex.com
6 All the complex networks metrics were extracted using the Gephi tool, available in https://gephi.org/
7 Available in https://elki-project.github.io/
8 See .