On technical trading and social media indicators for cryptocurrency price classification through deep learning

Ortu, Marco

;Uras, Nicola;Conversano, Claudio;Bartolucci, Silvia;Destefanis, Giuseppe

2022-01-01

---

Abstract

Predicting the prices of cryptocurrencies is a notoriously challenging task due to high volatility and new mechanisms characterising the crypto markets. In this work, we focus on the two major cryptocurrencies for market capitalisation at the time of the study, Ethereum and Bitcoin, for the period 2017–2020. We present a comprehensive analysis of the predictability of price movements comparing four different deep learning algorithms (Multi Layers Perceptron (MLP), Convolutional Neural Network (CNN), Long Short Term Memory (LSTM) neural network and Attention Long Short Term Memory (ALSTM)). We use three classes of features, considering a combination of technical (e.g. opening and closing prices), trading (e.g. moving averages) and social (e.g. users’ sentiment) indicators as input to our classification algorithm. We compare a restricted model composed of technical indicators only, and an unrestricted model including technical, trading and social media indicators. We found an increase in accuracy for the daily classification task from a range of 51%–55% for the restricted model to 67%–84% for the unrestricted one. This study demonstrates that including both trading and social media indicators yields a significant improvement in the prediction and accuracy consistently across all algorithms.