Prevendo a força de conexão por meio da rede social online facebook.

Abstract

In recent years the analysis of social networks has gained significant attention. Part of this success is due to the popularization of Online Social Networks and the large volume of data that these networks have generated. The work presents an algorithm to examine the Tie Strength between users of the online social network Facebook. In addition to the algorithm, the author developed a similarity metric to classify this Tie Strength. The similarity between a pair of users is based on their reactions to Facebook posts. The work methodology was based on the CRISP-DM model, a Data Mining process. Initially the work presents the processing for handling the database, especially publications, then the development of algorithms to measure the Tie Strength and similarity metric.Using the results generated by this algorithm, the work analyzes the existing connections between friends of the social network and shows that more than 90% of the profiles are connected to other similar profiles, indicating that the relationships built on Facebook are influenced by the interests of users. Finally, the work presents two clustering models, the Hierarchical Algorithm designed to build groups of varying sizes and the modified Hierarchical Algorithm developed by the author to build groups with the same number of profiles. The proposed algorithm demonstrated good results for clustering many profiles in a few groups.