Abstract
From a data analysis perspective, we propose a method called "node projection" to study the complex networks. The method projects network nodes into a high-dimensional measure space and uses the principal component analysis (PCA) to analyze the produced data cloud. By assigning similar nodes to close positions in the measure space, distinct topological patterns of random, regular, small-world and scale-free networks emerge. From the data analysis view, we also confirm the hierarchical structure and jellyfish model of the Internet Autonomous System fabric. We further use the node projection method to extract network feature based on PCA measurements and use them to quantitatively study similarity among selected networks. Similar networks are clustered together. The node projection method can be developed into a universal framework to discover network structural patterns on diverse scales.