Deep Learning and Sentiment Analysis-Based Cryptocurrency Price Prediction

Abstract

The rapid growth of Cryptocurrency dramatically influences the social and economic climate that has developed the trends for investors to seek opportunities for generating income from cryptocurrency investment trading. Cryptocurrency is volatile in nature due to the interdependence of cryptocurrency, market noise and many dependent factors. This has gained the attention of investors to rely on prediction models to forecast prices. Researchers proposed and implemented prediction models that utilized machine learning, deep learning algorithms and sentiment-based algorithm hybrid models. Researchers deduced that deep learning algorithms can capture the dependency features of cryptocurrency to increase accuracy in price prediction. In this paper, we proposed a system framework namely, DLCFS (Deep Learning Cryptocurrency Forecasting considering Sentiment), for cryptocurrency price prediction that considers the market features, trading volume, and interdependency between cryptocurrency and market sentiments. We conduct price forecasting for Bitcoin, Ethereum and Litecoin using their price history, and Reddit Submissions of cryptocurrency. Additionally, we have inferred the results for the performance of prediction models comparing DLCFS against machine learning. Results show that DLCFS outperformed the regression machine learning in predicting the price of Bitcoin, Litecoin, and Ethereum, considering market sentiment, with Correlation Coefficient (R) being 99.18%, 96.82% and 99.05% respectively.