Computing Signed Networks Structural Balance via Node Influenced Memetic Algorithm

The studies on structural balance of signed works have received a great attention due to its capability to describe the potential cooperation and conflicts among entities. Structure balance theory studies the unbalanced relationships in signed networks. The computation of structural balance aims to search for the least unbalance degree of a signed network to transform an unbalanced network into balanced one with the least cost. In this study, under the weak definition of structural balance theory, a node influenced memetic algorithm, called NIMA, is proposed to minimize the objective function. There are three main parts in NIMA. Firstly, a neighbor node influence-based initialization operation is applied to create an initial population for speeding the convergence process. Secondly, a node degree-based genetic operation is employed as the global search method. Moreover, a multi-level greedy local search is adopted to approach the potential optimum effectively. Extensive experiments on 9 real-world signed networks demonstrate that the proposed NIMA performs more efficiently, compared to other classic algorithms, on computing the structural balance of signed networks.