Financial markets have always been a point of interest for automated systems. Due to their complex nature, financial algorithms and fintech frameworks require vast amounts of data to accurately respond to market fluctuations. This data availability is tied to the daily market evolution so it is impossible to accelerate its acquisition. In this presentation, we discuss several solutions for augmenting financial datasets via synthesizing realistic time series with the help of generative models.

This problem is complex since financial time series present very specific properties, e.g., fat-tail distribution, the cross-correlation between different stocks, specific autocorrelation, cluster volatility, etc. In particular, solutions for capturing cross-correlations between different stocks and for transitioning from fixed to variable-length time-series without resorting to sequence modeling networks are presented.

A discussion is carried out on the problem of evaluating the quality of synthetic financial time-series. This presentation also introduces qualitative and quantitative metrics, along with a portfolio trend prediction framework, and discusses experiments carried out on real-world financial data extracted from the US stock market.