74. Conditional volatility modeling and financial risk measurement via recurrent neual networks
Invited abstract in session WC-6: Predictive Analytics: Forecasting I, stream Analytics, Data Science, and Forecasting.
Wednesday, 13:30-15:00Room: H9
Authors (first author is the speaker)
| 1. | Theo Berger
|
Abstract
We assess both simulated and empirical economic time series and discuss the application of recurrent neural networks to identify different volatility regimes of financial time series. We study statistical properties of the residuals to analyze the precision of competing concepts for conditional volatility modeling. Also, we demonstrate the application of financial risk measurement in comparison with econometric benchmarks.
The simulation assessment demonstrates that there exists a trade-off between insample accuracy and the detection of autocorrelation. We also introduce recurrent neural networks to a Value-at-Risk universe and demonstrate that recurrent neural networks describe a promising alternative to the econometric GARCH approaches for small training samples.
Keywords
- Forecasting
- Time Series Analysis
- Risk Analysis and Management
Status: accepted
Back to the list of papers