Study on the Qualitative Cohesion in Bitcoin Market Price Prediction

Abstract

Over time, various methodologies have been introduced for predicting the cryptocurrency market. While numerous studies have explored different variables, research incorporating the actual sentiments of investors has been scarce. In this study, we aimed to improve cryptocurrency market predictions by considering the qualitative cohesion. We built upon the existing LSTM model and extended our analysis to include RoBERTa and DistilBERT models through text mining. The results revealed that RoBERTa and DistilBERT incorporating investor sentiment outperformed the LSTM model in terms of prediction accuracy. Notably, the DistilBERT model, known for its exceptional word and context analysis, demonstrated the highest predictive power, followed by RoBERTa and the LSTM model. These findings underscore the importance of directly analyzing investor psychology in future market analyses. Furthermore, focusing on both individual words and contextual meaning is expected to yield even better market prediction results.

Keywords

• Cohesion
• cryptocurrency
• DistilBERT
• RoBERTa
• LSTM