基于多维属性的社会网络链接预测算法MDA-TF

Novel link prediction algorithm based on multi-dimensional attributes in social networks

针对传统社会网络链接预测算法忽视节点多维属性的问题,提出一种基于多维属性的社会网络链接预测算法MDA-TF。该算法首先经过数据预处理,结合节点的多维属性,构建张量模型;然后采用高阶正交迭代算法进行张量分解,得到核心矩阵和因子矩阵;最后根据核心矩阵生成链接预测结果。采用真实的社会网络数据集进行测试取得了较好的实验结果,实验结果也表明了该算法的有效性和正确性。

To address the problem of ignoring multi-dimensional attributes in social networks link prediction methods, this pa- per propsed a novel link prediction algorithm based on multi-dimensional attributes in social networks（ denoted as MDA-TF）. Firstly, after data preprocessing, combined with multi-dimensional attributes,it constructed a tensor model. Secondly it conduc- ted tensor factorization by high-order orthogonal iteration to get the core matrix and the factor matrix. Finally, it obtained the link prediction results according to the core matrix. The experimental results on the real social networks Show that the proposed algorithm achieves good prediction performance.