Study Of Influence Maximization Based On Network Embedding

The problem of influence maximization is a kind of problem specifically for the study on social networks.The so-called problem of influence maximization is to find out the k influence nodal sets in the network as the seed nodal sets,so as to maximize the influence propagation range generated by it.The data used in the study of influence maximization is all network data.It is known that network data is often sophisticated and thus it is challenging to deal with it.To process network data effectively,the first critical challenge is to find effective network data representation,and network embedding algorithm is a kind of highefficiency algorithm to process network data representation.In this work,we studied network embedding algorithm deeply,and proposed an improved DeepWalk algorithm named C-DeepWalk algorithm.Firstly,we proposed a new sampling method to generate node sequences.Then the generated node sequences were input into the Skip-Gram model for learning,and the vector representation of the node was obtained.Finally,by using link prediction task,experiments were carried out on several real data sets,and the performance of the C-DeepWalk algorithm proposed in this work was compared with existing network embedding algorithm.The results showed that the learning performance of C-DeepWalk algorithm is better than existing network embedding algorithm.At the same time,based on the study of the traditional heuristic influence maximization algorithm,this work introduced network embedded learning into the study of influence maximization,and proposed a new influence maximization algorithm named NEIM algorithm.Firstly,we combined C-DeepWalk algorithm and clustering algorithm to identified k clustering centers.Then,the KD tree algorithm was used to find the m nodes closest to each clustering center and take these nodes as candidate nodes.Finally,the greedy strategy was used to identify k influential seed nodes among the candidate seed nodes.From the two aspects of accuracy and efficiency,we have carried out a large number of experiments for the NEIM algorithm under different data sets and different propagation rates.The results showed that the efficiency of the NEIM algorithm is obviously higher than that of the greedy algorithm,and the accuracy is close to that of the greedy algorithm.