Abstract

In recent years, the cryptocurrency industry has experienced remarkable growth. Unlike traditional currencies, cryptocurrencies operate online without the need for a central authority, relying on cryptography to ensure secure and unique transactions. Despite the cryptographic safeguards in place, the cryptocurrency industry is still in its nascent stage, leading to questions about its potential applications. To gain a comprehensive understanding of public sentiment, this study focuses on analyzing sentiments related to Bitcoin. To achieve this, the research employs sentiment analysis and emotion recognition techniques by analyzing tweets related to digital currencies, a common approach used for predicting cryptocurrency values. The study introduces an ensemble model called LSTM-GRU, which evaluates the efficacy of a mixed LSTM and GRU recurrent neural network. Several ML and DL approaches, such as term frequency-inverse document frequency, word2vec, and the bag of words features, are investigated. Also included are emotion analysis models like Text Blob and Text2Emotion. Intriguingly, the results show that among other feelings, satisfaction over cryptocurrency acceptance ranks high. Using bag of words characteristics results in better performance for ML models, according to the study. Unbelievably, the proposed LSTM-GRU ensemble model outperforms state-of-the-art methods in both sentiment analysis (0.99) and emotion recognition (0.92). conventional machine learning methods and contemporary state-of-the-art models.