A General Framework for Portfolio Construction Based on Generative Models of Asset Returns

In this paper, we study a general framework for portfolio construction based on generative models of asset returns. We comprehensively investigate and compare portfolios built on simulations of several generative models. We also use various objective functions for portfolio construction and investigate portfolio blending techniques. Backtests are carried out on historical crypto-currency data from two major crypto-markets. We use fitted least absolute shrinkage and selection operator (LASSO) models for performance attribution. We show that the best performance can be achieved using vine-copula for the historical returns dependency model and maximizing the Sharpe ratio as the objective function. Blending or switching portfolios using L-inf similarity and the tanh activation function could further improve portfolio performances