Testing Different Reinforcement Learning Configurations for Financial Trading: Introduction and Applications

The construction of automatic Financial Trading Systems (FTSs) is a subject of research of high interest for both academic environment and financial one due to the potential promises by self-learning methodologies and by the increasing power of actual computers. In this paper we consider Reinforcement Learning (RL) type algorithms, that is algorithms that optimize their behavior in relation to the responses they get from the environment in which they operate, without the need for a supervisor. In particular, first we introduce the essential aspects of RL which are of interest for our purposes, then we present some original automatic FTSs based on differently configured RL algorithms and apply such FTSs to artificial and real time series of daily financial asset prices.