LSTM and GRU Neural Networks as a Support to Investment and Risk Management Strategies
Randomness dynamics in financial markets can be captured by the volatility, which directly links to the risk aversion of investors. As a consequence, a successful volatility prediction can be highly profitable from many aspects: For the investment manager, it can be a signal to hedge his investment portfolio in response to high upcoming volatility. For the option trader, it can be a signal for a volatility arbitrage. For the risk manager, it can be a support to forward-looking risk measures which depict the portfolio risk better.Two variations of Recurrent Neural Networks (RNN), the Long-Short Term Memory (LSTM) and the Gated Recurrent Units (GRU) models are employed to predict whether the volatility spread between the realized and implied volatility (RV-IV spread) ends up positive or negative on the next trading day. In particular, LSTM and GRU models share by construction the same structure design - prior tuning of hyperparameters, such that their prediction performances can be compared. SP500 Index and the Eurostoxx 50 Index are used as support for the classification prediction exercise anddesign of investment and risk management strategies. For both indices, the GRU model outperforms the LSTM model.Both investment and risk management strategies are built around the interpretation that a positive RV-IV spread prediction is indicative of a bear market to come as a result of higher upcoming volatility. The investment strategy, derived from SP500 GRU predictions, highlights that the GRU model does well in identifying a bear market and avoiding losses, albeit initially struggling to identify a bull market. The risk management strategy incorporates Eurostoxx 50 GRU predictions and machine learning techniques applied to volatilities and correlations to derive forward-looking risk measures in the case of an investment portfolio
Year of publication: |
[2023]
|
---|---|
Authors: | Guibert, Zacharie |
Publisher: |
[S.l.] : SSRN |
Subject: | Theorie | Theory | Neuronale Netze | Neural networks | Risikomanagement | Risk management |
Saved in:
freely available
Extent: | 1 Online-Ressource (14 p) |
---|---|
Type of publication: | Book / Working Paper |
Language: | English |
Notes: | Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments April 2023 erstellt |
Other identifiers: | 10.2139/ssrn.4428183 [DOI] |
Source: | ECONIS - Online Catalogue of the ZBW |
Persistent link: https://www.econbiz.de/10014354993
Saved in favorites
Similar items by subject
-
A machine learning approach to portfolio pricing and risk management for high-dimensional problems
Fernandez-Arjona, Lucio, (2020)
-
A machine learning approach to portfolio pricing and risk management for high-dimensional problems
Fernandez-Arjona, Lucio, (2022)
-
Managing weather risk with a neural network-based index insurance
Chen, Zhanhui, (2024)
- More ...