Portfolio optimization through hybrid deep learning and genetic algorithms vine Copula-GARCH-EVT-CVaR model

Abstract

This study investigates the potential benefits of using the Conditional Value at Risk (CVaR) portfolio optimization approach with a GARCH model, Extreme Value Theory (EVT), and Vine Copula to obtain the optimal allocation decision for a portfolio consisting of Bitcoin, gold, oil, and stock indices. First, we fit a suitable GARCH model to the return series for each asset, followed by employing the Generalized Pareto Distribution (GPD) to model the innovation tails. Next, we construct a Vine Copula-GARCH-EVT model to capture the interdependence structure between the assets. To refine risk assessment, we combine our model with a Monte Carlo simulation and Mean-CVaR model to optimize the portfolio. In addition, we utilize a novel version of deep machine learning's genetic algorithm to address the optimization decision. This research contributes new evidence to the CVaR portfolio optimization approach and provides insights for portfolio managers seeking to optimize multi-asset portfolios.