Deep Learning Approaches for High-Frequency Bitcoin Trading and Market Price Prediction

Abstract

The large price changes in Bitcoin have led to increased interest in predicting its future prices. This study uses a Deep learning approach to forecast Bitcoin values based on past data and technical indicators. Several models, including LSTM and Gradient Boosting, are used to find patterns in Bitcoin's price trends. Results show that the LSTM model reaches an accuracy of 98%. It also achieves a 97% accuracy rate for daily price predictions using logistic regression and linear discriminant analysis methods. The use of LSTM for predicting Bitcoin prices over time is more effective than traditional methods, especially regarding the importance of sample size in Deep learning. The analysis looks at Bitcoin price charts to project future changes, primarily through time series analysis that utilizes historical data. Deep learning algorithms are applied to identify complex patterns in Bitcoin pricing data. The dataset for these predictions is from KAGGLE, covering eight years from 2014 to 2022, which aids in developing more accurate forecasting techniques.