Design and Development of an Algorithmic Trading System for Cryptocurrencies using Reinforcement Learning and Deep Learning

Jaime Palos. (2020). Design and Development of an Algorithmic Trading System for Cryptocurrencies using Reinforcement Learning and Deep Learning. Trabajo Fin de Titulación (TFM). Universidad Politécnica de Madrid, ETSI Telecomunicación.

Abstract:
Machine Learning (ML) is growing and it’s rapidly making its way into almost every field. And the stock exchange and trading are no exception. Day after day, algorithmic trading is gaining more and more popularity. This technology relies on computer programs to execute algorithms that automate a trading strategy. We have been seeing more and more asset managers and hedge funds incorporate ML techniques to improve their trading strategies. And we have even seen computer-driven hedge funds that, according to The Financial Times, have already joined the industry’s top perform- ers [87]. Proof of this ML growth is that, nowadays, investment banks and hedge funds are em- ploying quantitative analysts to design and implement complex models that allow financial firms to price and trade securities, help with risk management, or even develop new trading strategies. And its importance can easily be seen reflected in their salaries; it is not uncom- mon to find positions with posted salaries of $250,000 or more and with bonuses, $500,000+ is achievable, according to Investopedia [116]. During the development of this Master’s Thesis, I will study the current state of these technologies and learn about, Stock, Trading, Machine Learning (ML), Algorithmic Trading, Reinforcement Learning (RL), Deep Learning (DL), etc. And apply this knowledge to design and develop an Algorithmic Trading System using different ML techniques. I will later compare the obtained results with the ones obtained in previous Algorithmic Trading projects and with the usual trading performance. I will end with a quick overview of possible improvements of the system or future lines of work.