Link Prediction:A Friend Recommendation Model

Recently, Link prediction in complex networks has attracted increasing attention from both physical and computer science communities. The research of link prediction mainly focuses on forecasting potential relations between nonadjacent nodes, including the prediction of the unknown links or the further nodes. An important fact in studying the link prediction is that the structural properties of networks have significant impacts on the performance of algorithms. Therefore, how to improve the performance of link prediction with the aid of structural properties of networks is an essential problem. To this, this paper is mainly divided into the following three aspects:(1).First of all, define the concept of weak clique structure; Then, a large number of real networks are analyzed and a common phenomenon is found:nodes are preferen-tially linked to the nodes with the weak clique structure(PWCS).(2) Based on this PWCS phenomenon, we propose a local friend recommendation (FR) index to facilitate link prediction. Our experiments show that the performance of FR index is generally better than some famous local similarity indices, such as Common Neighbor (CN) index, Adamic-Adar (AA) index and Resource Allocation (RA) index, and then analyzes the differences of FR index and RA index, gets the following conclusion:a high RA ranking value of links gives rise to a high FR ranking value. However, a high FR ranking value of links may induce a low RA ranking value of links. Finally, we want to know how the strength of PWCS affects the performance of FR index. For this purpose, we propose a generalized friend recommendation (GFR) index. Experiments show that:If the network has the phenomenon of PWCS, local community structure, strengthen the friend recommended model, the influence of link prediction effect will be better; If PWCS phenomenon more obvious, even reach maximum parameters, the influence of the characteristics of the weak clique structure is far insufficient, link prediction effect needs to be improved; If the network does not have PWCS phenomenon, link prediction effect will increase as the parameter. Based on the above results, according to whether the network has PWCS phenomenon and the phenomenon of PWCS is obvious design a better mixed friend recommendation model which can improve the accuracy of link prediction.(3) The friend recommendation model are generalized to weighted networks, the experiment shows that friend recommendation model on the weighted networks (WFR) is superior to WCN index, WAA index and WRA index, but also found WFR index prediction results are not as good as the unweighted networks has no right to FR index, analysis considering the weight may not be able to reflect the recommendation. In this paper, according to the number of common neighbors to redefine a false weights, and then applied to WFR index, experiments show that prediction effect is superior to FR index.