Autoencoding Conditional GAN for Portfolio Allocation Diversification
This is an incremental improvement for financial portfolio analysis, addressing limitations in existing generative models.
The paper tackled the problem of portfolio allocation by introducing an autoencoding conditional GAN (ACGAN) that learns historical trends and models market uncertainty, resulting in better portfolio allocation and more realistic generated series compared to Markowitz and CGAN methods.
Over the decades, the Markowitz framework has been used extensively in portfolio analysis though it puts too much emphasis on the analysis of the market uncertainty rather than on the trend prediction. While generative adversarial network (GAN) and conditional GAN (CGAN) have been explored to generate financial time series and extract features that can help portfolio analysis. The limitation of the CGAN framework stands in putting too much emphasis on generating series rather than keeping features that can help this generator. In this paper, we introduce an autoencoding CGAN (ACGAN) based on deep generative models that learns the internal trend of historical data while modeling market uncertainty and future trends. We evaluate the model on several real-world datasets from both the US and Europe markets, and show that the proposed ACGAN model leads to better portfolio allocation and generates series that are closer to true data compared to the existing Markowitz and CGAN approaches.