Toward random walk-based clustering of variable-order networks

Abstract Higher-order networks aim at improving the classical network representation of trajectories data as memory-less order $1$ Markov models. To do so, locations are associated with different representations or “memory nodes” representing indirect dependencies between visited places direct relations. One promising area investigation in this context is variable-order models it was suggested by Xu et al. that random walk-based mining tools can be directly applied on such networks. In paper, we focus clustering algorithms and show doing so leads to biases due number nodes each location. address them, introduce a aggregation algorithm produces smaller yet still accurate input sequences. We empirically compare found multiple real-world mobility datasets. As our model limited maximum $2$ , discuss further generalizations method higher orders.