Price Prediction of Digital Financial Assets: Using Machine Learning and Deep Learning

Abstract

Digitalisation of finance led to the creation of a digital financial economy, where digital assets such as cryptocurrencies, decentralized financial assets, non-fungible tokens, stablecoins, etc. were traded. In this study, machine learning and deep learning techniques, including ARIMA, FB Prophet, LSTM, and BiLSTM, have been used to forecast the prices of digital assets. In this study, Bitcoin, Ethereum, Uniswap, Aave, ApeCoin, and Decentraland tokens have been categorized into three groups, and the prediction models have been trained using the tokens' closing prices. The authors find that NFTs have been underestimated and that DeFi assets have greater growth potential. Whereas cryptocurrencies have been traded more and shown greater volatility than other asset classes. BiLSTM achieves the best results, with higher accuracy in price prediction. Here, it has been seen that ApeCoin, Decentraland, and Bitcoin are more stable than other assets. Thus, for an optimised portfolio and additional savings, it is necessary to provide a proper asset mix.