Extending the Adapted PageRank Algorithm centrality model for urban street networks using non-local random walks

In the urban street network domain, there is growing interest in extending conventional centrality measures to incorporate node-specific information (such as georeferenced socioeconomic data) in order to help identify important locations in an urban environment. One such centrality measure that is gaining attention is the Adapted PageRank Algorithm (APA) model. However, a fundamental concern with the APA model is the notion of teleportation because it means the random walker is equally likely to jump or ‘teleport’ to any intersection (node) in the street network, regardless of how far away it is. In this paper, we propose a centrality model that overcomes this counterintuitive idea. More specifically, we extend the APA model by modifying the jumping probabilities so that the random walker is more inclined to jump to a nearby intersection than a distant intersection. We accomplish this using non-local random walks which allow a random walker to jump to any node in the network with probabilities that depend on the distance separating the nodes. To demonstrate the differences between the two models, we present and discuss experimental results for a small ten node graph and a real-world urban street network.