Sentiment-Driven Price Prediction of the Bitcoin based on Statistical and Deep Learning Approaches

Abstract

Nowadays, Bitcoin has become the most popular cryptocurrency, which gains the attention of investors and speculators alike. Asset pricing is a risky and challenging activity that enchants lots of shareholders. Indeed, the difficulty in making predictions lies in understanding the multiple factors that affect the Bitcoin price trend. Modeling the market behavior and thus, the sentiment in the Bitcoin ecosystem provides an insight into the predictions of the Bitcoin price. While there are significant studies that investigate the token economics based on the Bitcoin network, limited research has been performed to analyze the network sentiment on the overall Bitcoin price. In this paper, we investigate the predictive power of network sentiments and explore statistical and deep-learning methods to predict Bitcoin future price. In particular, we analyze financial and sentiment features extracted from economic and crowd-sourced data respectively, and we show how the sentiment is the most significant factor in predicting Bitcoin market stocks. Next, we compare two models used for Bitcoin time-series predictions: the Auto-Regressive Integrated Moving Average with eXogenous input (ARIMAX) and the Recurrent Neural Network (RNN). We demonstrate that both models achieve optimal results on new predictions, with a mean squared error lower than 0.14%, due to the inclusion of the studied sentiment feature. Besides, since the ARIMAX achieves better predictions than the RNN, we also prove that, with just a linear model, we may obtain outstanding market forecasts in the Bitcoin scenario.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 63 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.