Reinforcement Learning for Finance

Part II. Data Augmentation

The second part of the book covers concepts about and approaches to generating data for financial deep Q-learning:

• Chapter 4 implements data generation approaches based on Monte Carlo simulation (MCS). One approach is to add white noise to an existing financial time series. Another one is to simulate financial time series data based on a financial model (a stochastic differential equation).

• Chapter 5 shows how to use generative adversarial networks (GANs) from AI, or more specifically, from deep learning (DL), to generate financial time series data that is consistent with and statistically indistinguishable from the target financial time series. Such a target time series can be the historical return series for a share of a company stock (think Apple shares) or historical foreign exchange quotes (think the EUR/USD exchange rate).