PageRank computation for Higher-Order Networks

TL;DR

This paper investigates PageRank computation on higher-order networks, identifies bias issues, and proposes an adapted model to improve ranking accuracy on real-world sequential data.

Contribution

It introduces a correction to PageRank for higher-order networks to address bias caused by item representation distribution.

Findings

Bias exists in PageRank due to item representation distribution

The adapted PageRank model reduces ranking bias

Application shows significant differences in rankings with the new method

Abstract

Higher-order networks are efficient representations of sequential data. Unlike the classic first-order network approach, they capture indirect dependencies between items composing the input sequences by the use of \textit{memory-nodes}. We focus in this study on the variable-order network model introduced in [Xu et al. (2016);Saebi et al. (2020)]. Authors suggested that random-walk-based mining tools can be directly applied to these networks. We discuss the case of the PageRank measure. We show the existence of a bias due to the distribution of the number of representations of the items. We propose an adaptation of the PageRank model in order to correct it. Application on real-world data shows important differences in the achieved rankings. \keywords{Higher-order Networks, Sequential data, Random walks, PageRank