A Novel Link Prediction Method for Multiplex Networks with Incomplete Information.

The study of network disruption has attracted much attention for its wide range of applications, including controlling the spread of epidemics, disrupting criminal networks, and controlling the spread of rumors, where the key is to find the key nodes in the network. Because the information is incomplete, the results of identifying key nodes are not accurate. In order to reduce the influence of incomplete information on network analysis, a link prediction algorithm based on multiplex networks' s characteristics is proposed. Multiplex networks will be divided into two parts: target layer and other layer, in which the target layer is incomplete information network. Each of other layers is assigned a weight, indicating how similar the network is to the target layer, while the edges of other layers were assigned a value, indicating their importance. The score of an edge is multiplied with the weight of the layer it is in, and the product of edges with the same endpoint is summed, and the result is used as the final score of edges with the same endpoint. Target layer will be supplemented by the selection of high-scoring edges. A multiplex network named Aarhus Computer Science Department was used. As a target layer, Facebook falls into three categories: complete information, incomplete information, and supplementary information. In each of these three cases, network collapse was performed. Experimental results show that this algorithm has higher AUC value and Precision than three classical link prediction algorithms based on common neighborhoods, preferential attachment and Jaccard index. In the case of complete information and supplementary information, the order of node deletion in network collapse is basically the same, which shows that this algorithm can effectively reduce the impact of incomplete information on network collapse.