A Cryptocurrency Price Prediction Model Based on Twitter Sentiment Indicators

Abstract

The cryptocurrency becoming increasingly expensive, price prediction methods have also been widely studied. As an application of big data in finance, the sentiment tendency of related topics on social platforms is an important indicator of cryptocurrency price prediction methods and has attracted broad attention. However, the accuracy of the existing macro-sentiment indicator calculation methods should be further improved. Aiming at the problem that the accuracy of price prediction is not significantly improved by applying the existing macro-sentiment indicators, this paper proposes three new public sentiment indicators based on small granularity. Correlational analysis between the indicators and price data is conducted in the paper as well. By analyzing the degree of sentiment tendency of each comment, the accuracy of the three public sentiment indicators is improved. Specifically, this paper quantifies public sentiment indicators by taking into account the degree of emotional bias of each tweet, which makes sentiment indicators with small granularity. Compared with previous methods, the value prediction accuracies of cryptocurrencies have been improved under three deep learning frameworks LSTM, CNN, and GRU with the use of small granularity sentiment indicators.