Predicting the variation of decentralised finance cryptocurrency prices using deep learning and a BiLSTM-LSTM based approach

Abstract

This study suggests a method for forecasting Decentralised Finance (DeFi). In order to represent DeFi cryptocurrencies, we have used DeFi Pulse Index (DPI) that tracks the performance of some of the largest protocols in the DeFi. To get prices and sentiment analysis from social media, we used the LunarCRUSH dataset. We then conducted a time series study of DPI price using a Bi-LSTM and Long Short-Term Memory (LSTM) model and compared it withwith LSTM, BiLSTM, and GRU models. The mean square error of our proposed model is 0.0005745, and the mean absolute error is 0.01891.