This repository contains the code for the thesis "The estimation of Value-at-Risk and Expected Shortfall based on deep generative models"

Author: Belonovskiy Peter Ilich, HSE DSBA

Supervisor: Naumenko Vladimir Vladimirovich, HSE Associate Professor

Thesis: https://www.hse.ru/en/edu/vkr/926006463?ysclid=m3e6gvwzru724833110

Installation

---

The dependency managament in project was implemented via poetry. To clone this repository and set up the environment, run the following commands:

git clone https://github.com/BELONOVSKII/var_es_dgm.git
cd var_es_dgm
poetry install
poetry shell

Note: poetry should be pre-installed in your system.

Download data

---

Thesis uses daily stock prices data from yahoo finance. To parse the yahoo finance and download data run:

python var_es_dgm/data_parcing/parse_yfinance.py 

This downloads individual stocks's data and produces combined file data/complete_stocks.csv that would be further used in the experiments.

Models

---

• Variance Covariance: var_es_dgm/basic_models/parametric.py
• Historical Simulation: var_es_dgm/basic_models/hist_sim.py
• TimeGrad: var_es_dgm/TimeGrad/

Experiments

---

• Univariate:experiments/univariate
• Multivariate: experiments/multivariate

Visualisations

---

All figures from the thesis could be created by running notebooks in visualisations/.