Community detection based on social interactions in a social network

Abstract

Recent research has involved identifying communities in networks. Traditional methods of community detection usually assume that the network's structural information is fully known, which is not the case in many practical networks. Moreover, most previous community detection algorithms do not differentiate multiple relationships between objects or persons in the real world. In this article, we propose a new approach that utilizes social interaction data (e.g., users' posts on Facebook) to address the community detection problem in Facebook and to find the multiple social groups of a Facebook user. Some advantages to our approach are (a) it does not depend on structural information, (b) it differentiates the various relationships that exist among friends, and (c) it can discover a target user's multiple communities. In the experiment, we detect the community distribution of Facebook users using the proposed method. The experiment shows that our method can achieve the result of having the average scores of Total-Community-Purity and Total-Cluster-Purity both at approximately 0.8.

Memory-based label propagation algorithm for community detection in social networks

“…We set limited capacity for each memory element is 10. The results of modularity measure for proposed algorithm as MLPA in comparison with other algorithm such as LPA [9], LPA-CNPE [16], LPA-CNP1 [16] and KLPA [15] are listed in table 4. figure 3, the results show a relative improvement in comparison with other community detection algorithms.…”

“…This heuristic was investigated over a maximization problem based on an adapted version of the clustering coefficient measure. In [9], researchers proposed a new method that utilizes social interaction data (e.g., users' posts on Facebook). Since many relationships are missing or not recorded, so they attempted to identify hidden communities.…”

No Reciprocity in "Liking" Photos

“…For example, [17] showed that, in social media, people form interlinked personal communities based on their follow and following connections as well as the norms, languages, and techniques used by them within the network. Somewhat differently, [9] argued that not all members are fully connected with each other and many relationships are missing in online social networks. They presented a new structure-based approach that leverages social communications (i.e., posts and replies) among users to identify different communities in which they engage.…”

Relationship Identification Across Heterogeneous Online Social Networks

“…The structure‐based researches are usually based on an assumption that people's different types of friends tend to form separate clusters, or in other words, social groups. Some researches addressed the relationship identification problem as a community detection problem using unsupervised methods (Chen et al ). And other works (Tang et al ) developed a supervised method utilizing network structure information and evaluated their proposed Relationship Classification Algorithm (RCA) on Renren (www.renren.com) social network.…”