Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating Credit VIX (CDS IV) Prediction Methods with Incremental Batch Learning

Robert Taylor

TL;DR

The paper tackles predicting CDS-implied volatility (Credit VIX) for the iTraxx Europe Main 1-Month index using three ML approaches under an incremental batch learning regime. It compares SVM, LightGBM, and an Attention-GRU hybrid, with a feature set inspired by Merton determinants and extensive feature engineering, including log-differences and realized volatility. Results show the Attention-GRU generally performs best, especially with longer training windows, but none decisively beats a naïve baseline according to the Diebold-Mariano tests, underscoring the challenge of out-of-sample IV forecasting in volatile regimes. The study highlights the value of hybrid attention-based architectures and incremental learning for online financial risk prediction, while also pointing to the critical role of feature engineering and the trade-offs in model complexity and computational cost for real-time deployment.

Abstract

This paper presents the experimental process and results of SVM, Gradient Boosting, and an Attention-GRU Hybrid model in predicting the Implied Volatility of rolled-over five-year spread contracts of credit default swaps (CDS) on European corporate debt during the quarter following mid-May '24, as represented by the iTraxx/Cboe Europe Main 1-Month Volatility Index (BP Volatility). The analysis employs a feature matrix inspired by Merton's determinants of default probability. Our comparative assessment aims to identify strengths in SOTA and classical machine learning methods for financial risk prediction

Evaluating Credit VIX (CDS IV) Prediction Methods with Incremental Batch Learning

TL;DR

The paper tackles predicting CDS-implied volatility (Credit VIX) for the iTraxx Europe Main 1-Month index using three ML approaches under an incremental batch learning regime. It compares SVM, LightGBM, and an Attention-GRU hybrid, with a feature set inspired by Merton determinants and extensive feature engineering, including log-differences and realized volatility. Results show the Attention-GRU generally performs best, especially with longer training windows, but none decisively beats a naïve baseline according to the Diebold-Mariano tests, underscoring the challenge of out-of-sample IV forecasting in volatile regimes. The study highlights the value of hybrid attention-based architectures and incremental learning for online financial risk prediction, while also pointing to the critical role of feature engineering and the trade-offs in model complexity and computational cost for real-time deployment.

Abstract

This paper presents the experimental process and results of SVM, Gradient Boosting, and an Attention-GRU Hybrid model in predicting the Implied Volatility of rolled-over five-year spread contracts of credit default swaps (CDS) on European corporate debt during the quarter following mid-May '24, as represented by the iTraxx/Cboe Europe Main 1-Month Volatility Index (BP Volatility). The analysis employs a feature matrix inspired by Merton's determinants of default probability. Our comparative assessment aims to identify strengths in SOTA and classical machine learning methods for financial risk prediction
Paper Structure (24 sections, 7 equations, 4 figures, 3 tables)

This paper contains 24 sections, 7 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Objectives
  3. Contributions
  4. Literature Review
  5. Existing Approaches in ML
  6. Methodology
  7. Data Description
  8. Candidate Models
  9. ATTN-GRU Architecture
  10. Feature Engineering
  11. Training and Validation
  12. Parameter State Configurations
  13. Experiments and Results
  14. Setup
  15. Evaluation Metrics
  16. ...and 9 more sections

Figures (4)

  • Figure 1: A high-level overview of the ATTN-GRU data flow, including residual connections.
  • Figure 2: Variance of Errors across ATTN-GRU, SVM and LightGBM, respectively.
  • Figure 3: Residuals of the forecasts (Unscaled - units are in log differences).
  • Figure 4: Predictions vs. Actuals (Scaled back to Index levels).