Machine learning for increased profits in the cryptocurrency market through pattern recognition with artificial neural networks

The cryptocurrency market is characterized by having a high price variation, being a developing market and by its uncertainty. Despite this, investors in this market are constantly making transactions which generate large amounts of data. Under these circumstances, whoever invests in this market is subject to high volatility, which allows both significant profits and exposure to high risk that can become big losses, and thus, these investors are looking to create strategies that maximize profits and minimize risks and operating costs. The complexity of the decisions around this problem makes it an attractive on for machine learning techniques. These exploit the amount of data to generate a predictive model, through pattern recognition, to support decision making. This work consists of a model for an investment strategy from both computational intelligence and financial information. The strategy aims for making investments that last 3 days, maximizes profitability and minimizes price volatility related risks, especially in moments with large and fast drops in prices. For this strategy, artificial neural networks were used, along with historical price data, data preprocessing and statistical indexes. The results were positive and showcased the possibility of achieving significant profits through this strategy during the testing period, superior to those achieved by the Buy and Hold market strategy.