Visit the website: Website Link
Trading is one of the most important aspects of commerce in human history.To consistently make profit, traders take every negative or positive trade they make as a learning opportunity. Predicting and analyzing financial indices has been of interest in the financial community for a long time, but recently, there has been a wide interest in the Machine Learning community to make use of and benefit from the more advanced techniques, such as reinforcement learning. Reinforcement learning (RL) has been able to surpass many of the Machine Learning benchmarks in a lot of fields. In this project we explore the plausibility of using RL in an algorithmic trading system. Our approach provides a feature rich environment for the reinforcement learning agent to work on. Firstly, we use the daily closing prices and volume of bitcoin in the market. As our aim is to provide long term profits to the user, we took into consideration some of the most reliable technical indicators. We have also developed a custom indicator to evaluate the current value of the prices. This aims at providing better insights of the Bitcoin market to the user. The Bitcoin market follows the emotions of the traders, so another factor of our trading environment was the overall daily Sentiment Score of the market on Twitter. The agent was tested for a period of 685 days which included the volatile period of Covid-19. It was able to provide recommendations reliable enough to give an average profit of 69%. Finally, the agent was able to recommend the optimal actions to the user through a website. The user can also access the visualizations of the important indicators to help them fortify their decisions.
Full Documentation with theory support: Report Link