Community detection in twitter based on tweets similarities in indonesian using cosine similarity and louvain algorithms

Abstract

Twitter is now considered as one of the fastest and most popular communication media and is often used to track current events or news. Many tweets tend to contain semantically identical information. When following an activity or news, sometimes in tweeting people do it in groups. Therefore, it is necessary to have a useful technique for grouping users based on the tweets similarities. In this study, cosine similarity method is used to examine the similarity of tweets between accounts, and a graph-based approach is proposed to detect communities. Graphs are first depicted from similarities between tweets and next community detection techniques are applied in graphs to group accounts that have similar tweets. The reason for using these two methods is that compared to other methods, the accuracy of cosine similarity is higher while Louvain can result a better modularity. From this research, it was concluded that cosine similarity and Louvain algorithm could be used in community detection on social media.