A Data Science Approach to Systemic Risk
summary
Simulated financial systems are generated with random graph generation and fed into an open-source C++ risk engine connected via a Python stack; the pipeline computes EEPE credit exposures for thousands of trades across 10 simulated systems with <50 banks each and shows about a 75% drop in average system risk under variation margin / initial margin rules. The prototype runs on a laptop, exposes per-bank inflow/outflow risk, and surfaces rare portfolios where collateralization increases exposure.
AI-written summary · about
chapters (16)
- 0:00Systemic risk and regulation intro
- 3:26Expert judgment limits
- 8:24Data science for systemic risk
- 12:02Simulation architecture
- 13:30Financial systems as graphs
- 15:40Risk in derivatives portfolios
- 17:48Risk graphs from trades
- 20:58Aggregating risk metrics
- 22:05Regulation impact evaluation
- 23:49Risk reduction results
- 25:15Regulation effects perspectives
- 27:00Data mining portfolios
- 28:44Simulation validation
- 29:26Future research directions
- 30:29Resources and collaboration
- 31:35Q&A
related talks
23mAn Open Source Architecture for Modernizing Market Risk Management
40mCreating a Contemporary Lending Risk Management System Using Python
47mCreating a Contemporary Risk Management System Using Python
31mRisk Analysis
39mMachine Learning in Financial Credit Risk Assessment
18mKeynote: AI for Financial Services
50mQuantifying Swiss Cheese, the Bayesian Way
28mThe COVID Scenario Pipeline: High Stakes Data Science