Machine Learning for Trading

The purpose of this paper is to discover whether it is possible to train a machine-learning algorithm to behave as a risk-adverse investor by using a dynamic model involving transaction costs. Transaction costs are frequently overlooked due to the complexity of integrating them into the learning algorithm used to train the trading system.

By specifying a series of mathematical and logical assumptions, it is concluded that machines can in fact learn an optimal total wealth function considering transaction costs while behaving like a risk-adverse investor. Total wealth in a portfolio can be identified at time t provided that a term for slippage is added into the equation to account for the fact that liquidating all open positions at the desired price is unlikely. Due to the complexity and non-linearity of a function of returns, the optimal-policy decision is best characterized using the mean-variance problem. It is assumed for the purposes of this study that the multivariate distribution p(r) is mean-variance equivalent.

Five components of reinforcement learning are explained as they are used in relation to creating a learning algorithm for trading—states, actions, value functions and policies, Q-learning, and the reward function. One limitation to this method is the need for millions of data observations for training steps, which may not be available or desirable. Therefore a simulation-based approach is introduced as a resolution, greatly simplifying this process. This paper contributes to the existing body of literature a way to estimate the cost function and optimal strategy by using only an asset return model. Read the full article by Gordon Ritter here.

Ready to test your ML strategy on your trading algo?

Get your free token to connect to the REST API and start trading. To generate your token:

  1. Register for a free practice account here.
  2. Log into and click on the account ID in the upper right corner.
  3. Click token management and generate your token.

Risk Warning: The FXCM Group does not guarantee accuracy and will not accept liability for any loss or damage which arise directly or indirectly from use of or reliance on information contained within the webinars. The FXCM Group may provide general commentary which is not intended as investment advice and must not be construed as such. FX/CFD trading carries a risk of losses in excess of your deposited funds and may not be suitable for all investors. Please ensure that you fully understand the risks involved.