Speaker: Marco Corazza, Università Ca' Foscari Venezia
The purpose of this talk is to design an intelligent stochastic control system for optimizing the performances of a financial trading system. Two model-free machine learning algorithms based on Reinforcement Learning algorithms are compared: Q-Learning and SARSA. Both these models optimize their behaviors in real time on the basis of the reactions they get from the environment in which operate. The idea to use Reinforcement Learning tools for designing intelligent trading systems is methodologically based on a new emerging theory about the financial market efficiency: the Adaptive Market Hypothesis. We develop both the models for trading on single stock price time series. They use simple state variables and operate selecting an action among three possible ones: buy, sell, and stay out from the market. We perform several applications based on different parameter settings which are checked on an artificial stock prices time series and on different daily real ones from the Italian stock market. Furthermore, performances are both gross and net of transaction costs.
Marco Corazza, PhD in Mathematics for the Analysis of Financial Markets, is associate professor at the Department of Economics at the Ca' Foscari University of Venice. Among his main research interests there are static and dynamic portfolio management theories, computational bio-inspired methodologies for optimization, multi-criteria methods for decision support. He participated (and participates) to several research projects. He is author of over one hundred scientific publications of national and international relevance, and provides consultancy activities.