Bitcoin Price Prediction Based on Sentiment of News Article and Market Data with LSTM Model

Chee Kean Chin, Nazlia Omar
dx.doi.org/10.17576/apjitm-2020-0901-01

Abstract

Bitcoin is a digital currency and investment tool that has received worldwide attention recently. However, the fluctuation of Bitcoin price has been a concern to the users and investors. Forecasting the Bitcoin price can serve as a guideline for investor and user to make effective strategy in their investment or usage. With the rapid development of the Internet, online data including news article can facilitate forecasting Bitcoin price. This research aims to study the effect of news article sentiment towards Bitcoin price with study period from September 2017 to August 2019. Accordingly, this study introduces sentiment analysis to understand the relevant information of online news articles and use it as an input feature for Bitcoin price prediction. Two main phases are included in the study, which is sentiment analysis and price prediction. In sentiment analysis, the sentiment is extracted based on a lexicon-based approach to capture the relevant news articles information regarding cryptocurrency markets. In price prediction, the sentiment is used as an input feature and Long Short-Term Memory (LSTM) model is used in price prediction phase. With Bitcoin market data and news articles as samples, the empirical results show that news articles sentiment reduced the overall error in Bitcoin price predictions.

keyword

Bitcoin, Prediction, LSTM, Lexicon, Sentiment

Area

Knowledge Technology