Dimension Reduction For Directed Network Data Based On Space-time Local Embedding

In the paper,we employed Space-time Local Embedding algorithm(Ke Sun,2015)to visualize the directed network(Dimension Reduction),and proposed an effective method of selecting the initial,and introduced local sparsification and two kinds of adaptive similarity algorithm to improve the visualization.As a qualitative analysis of data,visualization has no uniform quantitative evaluation index,therefore,we proposed a sequential clustering algorithm as the quantitative ranking evaluation index,and employed svm algorithm as the quantitative clustering evaluation index.According to the studies on toydata and the Journal Citation Network,we proved the algorithm is feasibile and proposed new ideas in the study on visualization of directed network.