Publication

Abstract : Cryptocurrency has taken the financial world by storm, with its value and relevance growing daily. For financial players, predicting cryptocurrency prices accurately has become crucial. Considering the growing importance of cryptocurrencies in the financial industry, this study focuses on the urgent problem of predicting their prices. Previous research in this domain has predominantly employed conventional statistical and machine learning methodologies, artificial neural networks, deep learning, and reinforcement learning. But lot of these methods often prove insufficient in capturing the complex patterns and sentiment analysis that are pivotal in influencing the dynamics of cryptocurrency markets. This research presents a one such method that captures these complex patters that is LSTM with self-attention mechanisms with Long Short-Term Memory (LSTM) neural networks to improve prediction accuracy and compares it with traditional LSTM. Results showcase the self-attention LSTM's ability to identify long-range dependencies, providing a robust solution for navigating the complexities of cryptocurrency markets. Furthermore, the self-attention LSTM demonstrates a notable improvement in R-squared (R2) scores over the traditional LSTM, underscoring its efficacy in capturing nuanced market dynamics and explaining more variance in the cryptocurrency price data.