Weighted Radial Variation for Node Feature Classification

02/23/2011 ∙ by C. Andris, et al. ∙ 0

Connections created from a node-edge matrix have been traditionally difficult to visualize and analyze because of the number of flows to be rendered in a limited feature or cartographic space. Because analyzing connectivity patterns is useful for understanding the complex dynamics of human and information flow that connect non-adjacent space, techniques that allow for visual data mining or static representations of system dynamics are a growing field of research. Here, we create a Weighted Radial Variation (WRV) technique to classify a set of nodes based on the configuration of their radially-emanating vector flows. Each entity's vector is syncopated in terms of cardinality, direction, length, and flow magnitude. The WRV process unravels each star-like entity's individual flow vectors on a 0-360 spectrum, to form a unique signal whose distribution depends on the flow presence at each step around the entity, and is further characterized by flow distance and magnitude. The signals are processed with an unsupervised classification method that clusters entities with similar signatures in order to provide a typology for each node in the system of spatial flows. We use a case study of U.S. county-to-county human incoming and outgoing migration data to test our method.



There are no comments yet.


This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.