Modelling and prediction of financial trading networks: An application to the NYMEX natural gas futures market

TL;DR

This paper introduces a new statistical modeling approach using stochastic blockmodels and hidden Markov models to analyze and predict the structure of financial trading networks, demonstrated on NYMEX natural gas futures data.

Contribution

It develops a novel methodology combining stochastic blockmodels and hidden Markov models for dynamic analysis and prediction of financial trading networks.

Findings

Effective short-term prediction of market transactions.

Identification of market-structuring events.

Application to NYMEX natural gas futures data.

Abstract

Over the last few years there has been a growing interest in using financial trading networks to understand the microstructure of financial markets. Most of the methodologies developed so far for this purpose have been based on the study of descriptive summaries of the networks such as the average node degree and the clustering coefficient. In contrast, this paper develops novel statistical methods for modeling sequences of financial trading networks. Our approach uses a stochastic blockmodel to describe the structure of the network during each period, and then links multiple time periods using a hidden Markov model. This structure allows us to identify events that affect the structure of the market and make accurate short-term prediction of future transactions. The methodology is illustrated using data from the NYMEX natural gas futures market from January 2005 to December 2008.