社会网络中基于核函数的信息传播模型

摘 要:传统的社会网络信息传播模型假设整个网络结构是已知的，并在已知的网络上分析信息的动态传播，然而实际的社会网络往往是不完全的。为了研究信息在不完全社会网络上的信息传播，本文提出了一种基于核函数的信息传播模型。首先，根据信息传播在社会网络中的级联关系将网络中的节点映射为到连续的特征空间，并通过节点间的距离反应节点的传播顺序。其次，将信息在网络中的传播描述为特征空间中的能量扩散过程，并采用随机梯度下降法进行优化求解。最后，将信息的内容加入到目标特征空间中，并给出了相应的核函数。实验表明，本文提出的信息传播模型与相关的模型相比较不仅可以弥补社会网络的不完全性，还具有更高的预测性能。

Kernel Function Based Information Diffusion Model in Social Networks

Abstract: Traditional information diffusion models usually assume that the underlying social network is known, and analyze the dynamic diffusion of information on the known network, but real social networks are usually uncompleted. In order to analyze the diffusion of information on uncompleted social networks, this paper proposes a kernel based information diffusion model. Firstly, we map users in information cascades to latent continuous feature space, and represent the orders of users in a cascade as distances between points in feature space. Secondly, we describe the diffusion of information with energy diffusion process in feature space, and solve the optimization problem with stochastic gradient decent algorithm. Finally, we take the content of information into consideration, and give the corresponding kernel function. The experiments show that, the proposed model can handle uncompleted social networks, and has better performance than related works.