Elucidating Nuances

Abstract

Cryptocurrencies are characterized by their volatility and growth. This necessitates effective price prediction methods. Machine learning algorithms have emerged to be effective, particularly long short-term memory (LSTM) models, in price prediction. This study, therefore, compares the LSTM models, namely LSTM, BiLSTM, and CNN-BiLSTM, to identify the superior LSTM model, along with the Crypto fear and greed index (FGI) for price prediction. The study revealed the difference in the performance of different LSTM algorithms in price prediction. While LSTM and BiLSTM showed strong predictive power, CNN-BiLSTM showed slightly varied performance across cryptocurrencies. The FGI also had a mixed influence based on the specific cryptocurrency. The study reinforces LSTM's effectiveness in cryptocurrency price prediction and the need for careful understanding when integrating sentiment data like the FGI. The findings have implications for investors and researchers seeking to leverage LSTM algorithms and sentiment indicators to make informed decisions in these market dynamics.