Estimating value at risk: LSTM vs. GARCH
Ormaniec, Weronika, Pitera, Marcin, Safarveisi, Sajad, Schmidt, Thorsten
Estimating value-at-risk on time series data with possibly heteroscedastic dynamics is a highly challenging task. Typically, we face a small data problem in combination with a high degree of non-linearity, causing difficulties for both classical and machine-learning estimation algorithms. In this paper, we propose a novel value-at-risk estimator using a long short-term memory (LSTM) neural network and compare its performance to benchmark GARCH estimators. Our results indicate that even for a relatively short time series, the LSTM could be used to refine or monitor risk estimation processes and correctly identify the underlying risk dynamics in a non-parametric fashion. We evaluate the estimator on both simulated and market data with a focus on heteroscedasticity, finding that LSTM exhibits a similar performance to GARCH estimators on simulated data, whereas on real market data it is more sensitive towards increasing or decreasing volatility and outperforms all existing estimators of value-at-risk in terms of exception rate and mean quantile score.
Jul-21-2022
- Country:
- Europe
- Germany (0.28)
- United Kingdom (0.46)
- Europe
- Genre:
- Research Report > New Finding (0.34)
- Industry:
- Banking & Finance > Trading (1.00)
- Technology: