Sentiment Analysis for Bitcoin Prediction

Since its first introduction by Nakamoto (2008), Bitcoin has grown to become the lead- ing peer-to-peer blockchain-based cryptocurrency in the world by market capitalisation. With the increasing amount of capital invested in Bitcoin, the effort to predict the price of the currency increased as well (Corbet, Lucey, Urquhart, & Yarovaya, 2019). Due to the novelty of the entire cryptocurrency asset class and its decentralized management, it is difficult to assess Bitcoins fundamental value (Jiang, Nie, & Ruan, 2018). Neverthe- less, both Urquhart (2016) and Nadarajah and Chu (2017) rejected the Efficient Marktet Hypothesis for Bitcoin, indicating that Bitcoin is indeed predictable.

We present the predictive performance of Stockpulse’s sentiment data for the Bitcoin price.

Stockpulse’s sentiment data reaches up to 90% correlation with the Bitcoin price on a 0.99 significance level.

In a time frame from 01.01.2021 until 01.06.2022, our sentiment- based LSTM model predicts the daily price with a mean absolute percentage error of 3.55%.

Request Paper