Optimization of Cryptocurrency Predictions Using a Deep Learning Approach

Abstract

Cryptocurrency is a decentralized digital currency that a central government regulates. Since cryptocurrencies are highly volatile, analysis is required before using cryptocurrencies to minimize losses. This research compares the Long Short Term Memory (LSTM) model and optimization algorithms such as Adam and Root Mean Square Propagation (RMSProp) to predict cryptocurrency values. The LSTM method was optimized using the Adam Optimizer and evaluated based on the Root Mean Square Error (RMSE). Thus the predicted RMSE value is 0.08217562639465784, which is a slight error value so that it is close to the actual value. While the RMSE value of 0.10699215580552895 using RMSProp gets a more significant value which impacts the accuracy of the prediction results. Thus the combination of the LSTM and Adam algorithms can accurately predict and optimize data.