Embedding and correlation tensor for XRP transaction networks

TL;DR

This paper analyzes XRP transaction networks from 2017-2018 by embedding weekly networks into vector space, constructing a correlation tensor, and applying double SVD to reveal systemic insights, especially during market price bursts.

Contribution

It introduces a novel approach combining network embedding and correlation tensor analysis to study XRP transaction networks and market dynamics.

Findings

Correlation tensor captures structural regularities in XRP networks.

Double SVD reveals key systemic features during market bursts.

Model parameters significantly influence correlation tensor properties.

Abstract

Cryptoassets are growing rapidly worldwide. One of the large cap cryptoassets is XRP. In this article, we focus on analyzing transaction data for the 2017-2018 period that consist one of the significant XRP market price bursts. We construct weekly weighted directed networks of XRP transactions. These weekly networks are embedded on continuous vector space using a network embedding technique that encodes structural regularities present in the network structure in terms of node vectors. Using a suitable time window we calculate a correlation tensor. A double singular value decomposition of the correlation tensor provides key insights about the system. The significance of the correlation tensor is captured using a randomized correlation tensor. We present a detailed dependence of correlation tensor on model parameters.