融合级联活跃转发者的社交网络影响最大化方法

已有的社交网络影响最大化研究大多基于网络图结构，没有利用级联数据蕴涵的信息，不能有效捕捉用户间的真实影响。针对此问题，基于级联数据提出融合级联活跃转发者的影响最大化方法。该方法首先设计融合活跃转发者的嵌入式神经网络模型，通过级联转发记录有监督地训练获取用户特征向量，然后根据信息可达对象数量和扩散概率计算融合活跃转发者的用户影响力，最后基于贪婪策略选择种子集。在三种大规模数据集上同四种具有代表性的方法进行对比，实验结果表明，在信息真实扩散范围方面，提出的方法更具有效性。

Integrating active information forwarder of cascade for influence maximization

The existing works for solving the problem of influence maximization mainly focused on graph and didn''t fully exploited the information cascade, which couldn''t effectively capture the actual influence between users. To this end, this paper proposed a new approach integrating active information forwarder for influence maximization based on the information cascade. Firstly, this approach designed an embedded neural network model considering active information forwarder and obtained the feature vectors of users by training model supervised by the actual diffusion record of the cascade data. Then, it presented a measurement method of user influence combining the feature vector of active information forwarder and information initiator according to the number of information reachable objects and diffusion probability. Finally, it selected the seeds by using greedy policy. The experimental results with four approaches on three large-scale data sets show the validity of this proposed approach in the actual spread of information.