News reports in social media are presented with large volumes of different kinds of documents. The presented topics in these documents focus on different communities and person stances and opinions. Knowing the relationships among persons in the documents can help the r More
News reports in social media are presented with large volumes of different kinds of documents. The presented topics in these documents focus on different communities and person stances and opinions. Knowing the relationships among persons in the documents can help the readers to obtain a basic knowledge about the subject and the purpose of various documents. In the present paper, we introduce a method for detecting communities that includes the persons with the same stances and ideas. To do this, the persons referenced in different documents are clustered into communities that have related positions and stances. In the presented method. Community-based personalities are identified based on a friendship network as a base method. Then by using a genetic algorithm, the way that these communities are identified is improved. The criterion in the tests is rand index of detection of these communities. The experiments are designed based on real databases that published in Google News on a particular topic. The results indicate the efficiency and desirability of the proposed method
Manuscript profile
Community detection in temporal social networks is one of the most important topics of research which attract many researchers around the world. There are variety of approaches in detecting communities in dynamic social network among which label propagation approach is More
Community detection in temporal social networks is one of the most important topics of research which attract many researchers around the world. There are variety of approaches in detecting communities in dynamic social network among which label propagation approach is simple and fast approach. This approach consists of many methods such as LabelRankT is one with high speed and less complexity. Similar to most methods for detecting communities in dynamic social networks, this one is not trouble free. That is, it is not considered the internal connection of communities, when it expands communities of the previous snapshots in the current snapshot. This drawback decreases the accuracy of community detection in dynamic social networks. For solving the drawback, a greedy approach based on local modularity optimization is added to LabelRankT method. Here, the newly proposed GreedyLabelRankT, LabelRankT and non-overlapping version of Dominant Label Propagation Algorithm Evolutionary (DLPAE-Non Overlapping) on real and synthetic datasets are implemented. Experimental results on both real and synthetic network show that the proposed method detect communities more accurately compared to the benchmark methods. Moreover, the finding here show that running time of the proposed method is close to LabelRankT. Therefore, the proposed method increase the accuracy of community detection in dynamic social networks with no noticeable change in the running time of that.
Manuscript profile
