Twitter Sentiment Analysis-Based Adjustment of Cryptocurrency Action Recommendation Model for Profit Maximization

Cryptocurrencies have recently attracted considerable attention, resulting in research mainly on deep learning-based price prediction models to maximize profit. Two approaches been adopted. Studies adopting the first approach directly predict future cryptocurrency price. Long short-term memory (LSTM) and gated recurrent unit (GRU), which show high performance time-series data, are used for this approach. Further, studies second recommend actions investors profits, such as “Sell”, “Buy”, “Wait.” In approach, classification results derived based probabilities. However, these action recommendation do not consider quality of result. For example, it is risky accept result when probability that model two classes correct answer approximately 51%. To solve problem, we a method adjusting Twitter sentiment analysis. The experimental proposed adjustment improves by 3% compared conventional methods statistically validated.