Extract the Backbone of Global Value Chain

Network Pruning is a basic step in high-dimensional network data processing, helping researchers better understand the topological characteristics and grasp the key points of analysis. Among numerous existing methods, we believe there is no appropriate one for the ICIO network, which is almost a both directed and weighted complete multigraph even with all the self-loops that could exist, and thus propose two principles under the perspective of econophysics: on one hand, the information content of network should be retained as much as possible with the decline of number of edges; on the other hand, only the truly important edges linking nodes are worthy to be retained during downsizing the network. To find out a trade-off between them, we design the SPFE method that is substantially the combination of two methods, which are based on searching pivotal paths and filtering redundant edges, respectively. In sum, the SPFE-based backbone of GVC will pave the way for the ICIO network’s community detection, link prediction, and spatial econometrics, etc.