Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Conditional Forecasting of Margin Calls using Dynamic Graph Neural Networks

Matteo Citterio, Marco D'Errico, Gabriele Visentin

TL;DR

This work shows that the network dynamics can be successfully incorporated into stress-testing practices, thus providing regulators and policymakers with a crucial tool for systemic risk monitoring.

Abstract

We introduce a novel Dynamic Graph Neural Network (DGNN) architecture for solving conditional $m$-steps ahead forecasting problems in temporal financial networks. The proposed DGNN is validated on simulated data from a temporal financial network model capturing stylized features of Interest Rate Swaps (IRSs) transaction networks, where financial entities trade swap contracts dynamically and the network topology evolves conditionally on a reference rate. The proposed model is able to produce accurate conditional forecasts of net variation margins up to a $21$-day horizon by leveraging conditional information under pre-determined stress test scenarios. Our work shows that the network dynamics can be successfully incorporated into stress-testing practices, thus providing regulators and policymakers with a crucial tool for systemic risk monitoring.

Conditional Forecasting of Margin Calls using Dynamic Graph Neural Networks

TL;DR

This work shows that the network dynamics can be successfully incorporated into stress-testing practices, thus providing regulators and policymakers with a crucial tool for systemic risk monitoring.

Abstract

We introduce a novel Dynamic Graph Neural Network (DGNN) architecture for solving conditional -steps ahead forecasting problems in temporal financial networks. The proposed DGNN is validated on simulated data from a temporal financial network model capturing stylized features of Interest Rate Swaps (IRSs) transaction networks, where financial entities trade swap contracts dynamically and the network topology evolves conditionally on a reference rate. The proposed model is able to produce accurate conditional forecasts of net variation margins up to a -day horizon by leveraging conditional information under pre-determined stress test scenarios. Our work shows that the network dynamics can be successfully incorporated into stress-testing practices, thus providing regulators and policymakers with a crucial tool for systemic risk monitoring.

Paper Structure

This paper contains 32 sections, 34 equations, 27 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background on Dynamic Graph Neural Networks (DGNNs)
  3. Encoder-decoder framework and model training.
  4. GC-LSTM.
  5. EvolveGCN
  6. Conditional forecasts.
  7. Multi-step ahead prediction.
  8. Problem description
  9. Overnight reference rate.
  10. Overnight Indexed Swaps (OIS).
  11. Dynamic financial network model.
  12. Variation margin forecasting problem.
  13. Proposed DGNN model
  14. GNN module
  15. Pricing module
  16. ...and 17 more sections

Figures (27)

  • Figure 1: Learning framework of a typical discrete-time DGNN.
  • Figure 2: Encoder-decoder framework for DGNNs.
  • Figure 3: Data windowing process for a discrete-time DGNN.
  • Figure 4: Conditional forecasting within the encoder-decoder DGNN framework.
  • Figure 5: Seq2Seq setup for $m$-steps ahead prediction.
  • ...and 22 more figures

Theorems & Definitions (1)

  • Definition A.1: CIR model