An Improved New Link Prediction Method in Social Multilplex Networks Based on the Gravitational Search Algorithm

The analysis of large scale dynamic networks provides useful information for the network administrator. This plays an important role in modern societies. The prediction of missing links or possible links in the future is an important and interesting issue on social networks that can support important applications with features such as new recommendations for users, friendship suggestions, and discovery of forged connections. Many real-world social networks display communications in multi-layers (for example, several social networking platforms). In this research, the problem of link prediction in multiple networks has been  studied and a new link prediction method in multiplex networks, based on unsupervized graph structure and the gravitational search algorithms is presented. Different layers of the multiplex network have been used to increase the accuracy of the proposed method and we have presented a methodology that uses information from other layers and community information where people are associated. We have provided this information in the form of a rating. These privileges, in a way, determine the prediction of the edges between individuals in these types of networks. One of the criteria for comparing predictive algorithms is to calculate the AUC for these algorithms and using this criterion for comparison accompanied by a travian data set used as a benchmark, it is seen that the AUC of our method has improved 7% compared to Adamic which is a similar method. The results demonstrate that using community information and the gravitational algorithm in layered networks improves link prediction.