The Modified Gravitation-based Algorithm for Community Detecting in Dynamic Social Networks

With in-depth research on the complex network, people have a deeper understanding of the physical significance and mathematical characteristics of social networks. Communities are clusters of closely connected nodes within a social network. Community detection is an important task to understand the characteristics of social networks. The gravitation-based algorithm (GBA) is proposed in our previous work to simulate the process of community evolution based on Newton's law of universal gravitation. Based on GBA, we present a modified gravitation-based algorithm (MGBA) in this paper. MGBA uses the degree of nodes in the network to represent the mass of nodes, redefines the distance parameter in the law of gravity according to the close structural characteristics of the community, and introduces the concepts of the parent node and the center node. MGBA includes the MAGB algorithm and the MAFMG algorithm. MAGB is used to detect the community structure and find the center community. MAFMG algorithm is used to find the max gravity of the community. On this basis of MGBA, we present a dynamic modified gravitation-based algorithm (DMGBA) which is based on incremental clustering. After using MGBA to detect the community structure in static networks. If edges are added or removed, DMGBA will be used to redivide the dynamical communities. For adding a new edge, add the neighbor nodes of two related nodes, modify the quality of each node, and then put these nodes into a collection. For deleting old edges, delete the neighbor nodes of two related nodes, and modify the mass of each node. Change the central node of related nodes to itself, and put them into the collection if they are still in the network. The experimental results show that MGBA algorithm achieves 1% higher F1-score in less time than GBA in static large-scale social networks, DMGBA can get good partitions of communities in much less time than comparison methods.