Analysis of Technical, Economic and Social Information Features to Predict the Bitcoin Price Direction for Day-Trade Operations

Cryptocurrencies are one of the most important financial and technological innovations of recent years. Currently, the interest in Bitcoin has grown, for traders and the general public. However, its high volatility represents a challenge in terms of prediction models for day-trade operations. In this way, recent studies have been proposed to predict the Bitcoin price direction for day-trade operations, but the maximum accuracy obtained was around 57.5%. In order to contribute and advance the state-of-the-art, this article experiences the impact of use Blockchain data, international economic indices, social trends information and technical indicators to overcome the predictions of the Bitcoin price direction. Thus, it is proposed a methodology based on data collection and processing, where weighted moving averages (for the Blockchain data, economic indices and social media trends) and technical indicators (considering the Bitcoin exchange rate data) were extracted/calculated from the original databases. The attributes were submitted to the Information Gain algorithm to select the most relevant ones. In the sequence, Support Vector Machines and Artificial Neural Networks models were used to predict the Bitcoin price direction for day-trade purposes. As a result, it was possible to obtain an average accuracy of 63.84% in a 1-year prediction period, overcoming other related studies.