Community Overlapping Detection Algorithm based on Bias-Subordination

摘要

The real-world network structure often presents the characteristics of overlapping communities. High-quality communities are helpful to understand the real complex networks. The discovery of overlapping communities has become a hot spot in current recommendation research algorithms. Regarding the randomness and instability problems in the original overlapping community discovery algorithm, this paper proposes the overlapping community discovery algorithm CODA-BS, which is based on the priority of community degree subordination. In each iteration process, the CODA-BS algorithm uses information entropy to calculate the vertex threshold, uses the weighted function proposed in this paper to optimize the label membership coefficient, and performs label screening and normalization of the membership coefficient according to the label propagation rules to detect a more ideal community. Tests on the benchmark data set and data sets of Chinese Association for Science and Technology scholars, show that the CODA-BS algorithm has higher accuracy and improved stability than COPRA algorithm, and the CODA-BS algorithm is more accurate and more stable than the classic COPRA algorithm.