Learning credit networks with applications in financial industry
summary
A concrete method learns a Bayesian credit network from CDS spreads by detecting epsilon drops (thresholded jump points) and fitting structure and parameters with bnlearn, then visualizing and analyzing propagation with NetworkX. The pipeline produces conditional default probabilities that plug into CVA pricing and shows crises make hubs far more contagious; a full CVA run requires about 3.75 billion valuations so GPU acceleration is applied to make it tractable.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 3:15Lehman collapse global impact
- 6:35Systemic risk complexity
- 9:24Derivatives basics
- 13:44Counterparty risk CVA
- 15:54Measuring co-dependencies
- 19:10Russian crisis case study
- 21:17Bayesian networks risk
- 25:12CDS spreads distress signals
- 27:02Calibrating calm vs distress
- 29:32Stress propagation
- 30:37Systemic risk management
- 31:01Q&A Bayesian assumptions
- 32:04Q&A Risk propagation steps
- 33:28Q&A Time series analysis
- 34:10Q&A Trading dependencies
- 34:56Q&A Network transformations
related talks
39mMachine Learning in Financial Credit Risk Assessment
34mMaximizing Credit Card Profitability with ML–Driven CLV
18mTransforming Financial Decision Intelligence Through Advanced NNs
48mProcessing Terabytes of Data in Credit Scoring
36mA Data Science Approach to Systemic Risk
14mCloud-Native AI for Fraud Prevention with Bayesian Networks
31mLet's do graph machine learning on Cloud
41mBayesian Network Modeling using R and Python