Turkish Journal of Electrical Engineering and Computer Sciences




A community is a group of people that shares something in common. The definition of the community can be generalized as things that have common properties. By using this definition, community detection can be used to solve different problems in various areas. In this study, we propose a new network-based community detection algorithm that can work on different types of datasets. The proposed algorithm works on unweighted graphs and determines the weight by using cosine similarity. We apply a bottom-up approach and find the disjoint communities. First, we accept each node as an independent community. Then, the merging process is applied by using the modularity value as a stopping criterion. We use real datasets and evaluate the algorithm with modularity, normalized mutual information, and performance metrics. In addition, we test our algorithm by using central nodes. We also take into consideration the number of communities in the case they are known. The proposed algorithm has high modularity and accuracy in different datasets.


Community detection, disjoint communities, cosine similarity, modularity, network-based

First Page


Last Page