Bitcoin Prediction using Convolutional Neural Network

Abstract

This research study addresses the challenging issue of predicting the highly volatile price of Bitcoin, which exhibits significant fluctuations. Surprisingly, despite their successful applications in various engineering and scientific domains, Convolutional Neural Networks (CNNs) have been largely neglected in the context of financial time series modeling. Therefore, the primary emphasis of this study is on exploring the potential of CNNs for improving Bitcoin price prediction. Our model harnesses the capabilities of CNN to extract crucial features from historical Bitcoin transaction data, revealing underlying patterns and trends. Subsequently, LSTM is employed to analyze and predict Bitcoin price movements based on these learned features. To bolster prediction accuracy, we integrate external factors, including macroeconomic variables and investor sentiments, with the historical Bitcoin data. The model’s performance is rigorously evaluated on a comprehensive real-world dataset, assessing its effectiveness in predicting Bitcoin’s price direction and magnitude. The results demonstrate promising prediction accuracy, surpassing conventional models and showcasing the potential of leveraging artificial intelligence in cryptocurrency price forecasting.