Overlapping community detection based on contribution value improved SLPA

Overlapping community detection algorithms play an important role in complex network analysis. The SLPA algorithm is an extended algorithm for the Label Propagation Algorithm (LPA), which makes the original algorithm transition from only exploring non-overlapping communities to overlapping communities. However, due to the randomness of its initialization phase and label update phase, the SLPA algorithm has fatal disadvantages the instability of the partitioning results and the quality of the partition cannot be guaranteed. This paper proposes an improved algorithm for SLPA algorithm, which uses the index of degree centrality to reconstruct the initialization sequence. At the same time, we proposes a concept of contribution value to rewrite the update rule, eliminating the randomness of the phase of initialization, and updating the label value by the contribution value. Making dense areas are more compact and loose areas are more sparse in the network, and eventually achieving detection of overlapping communities. The experimental results show that the CSLPA algorithm performs better in artificial networks and real networks than the existing classical algorithms mentioned in this paper.