Financial Trading Systems: Is Recurrent Reinforcement Learning the Way?

Abstract

In this chapter we propose a financial trading system whose trading strategy is developed by means of an artificial neural network approach based on a learning algorithm of the recurrent reinforcement type. In general terms, this kind of approach consists, first, of directly specifying a trading policy based on some predetermined investor’s measure of profitability, and second, of directly setting the financial trading system while using it. In particular, with respect to the prominent literature, in this contribution we take into account as a measure of profitability the reciprocal of the returns weighted direction symmetry index instead of the widespread Sharpe ratio, and we obtain the differential version of the measure of profitability we consider and all the related learning relationships. Finally, we propose a simple procedure for the management of drawdown-like phenomena, and finally, we apply our financial trading approach to some of the most prominent assets of the Italian stock market.