The Research about the Link Clustering Algorithmic Based on Tensor Analysis

The multi-link high-dimensional data clustering problem of complex information networks is difficult to handle and less efficient. Tensor is a mathematical representation of high-dimensional data (multidimensional array), Tensor decomposition technique can convert the high-dimensional data into the low-dimension data effectively and the more and more applied in the field of data mining, such as Tucker, CP, HOSVD, PARAFAC decomposition method so on. The modularity network analysis method can be used to the spectrum clustering. So we propose a novel link clustering algorithm based on higher order tensor analysis methods and modularity network analysis. Using modularity approach to analysis networks, using multi-dimensional tensor expressed in the form of complex multi-link data, using Tucker tensor decomposition method to reduce the dimensions of the data and the time and space complexity of the algorithm. The effectiveness and robustness of the algorithm is tested in complex network environment and proved that is better than the kindred algorithms.