Cryptocurrency price predictor

Abstract

Cryptocurrency is a recent innovation in the financial sector that takes investment and research into account. Cryptocurrencies are difficult to predict in terms of value due to their rapid value declines over brief periods of time. At the moment, more than a thousand cryptocurrencies are registered and in use. This article examines the use of recurrent neural network algorithms to the prediction of cryptocurrency values, with particular attention on Bitcoin, Ethereum, and Litecoin. In particular, this study looks at three RNN algorithms: biLSTM, LSTM, and gated recurrent unit (GRU)—all based on mean absolute percentage error (MAPE). estimating the price of bitcoin on the market. The results show that the GRU algorithm outperforms the other two algorithms for BTC, LTC, and ETH in terms of accuracy and MAPE percentages of 0.2116%, 0.2454%, and 0.8267%. Conversely, the biLSTM algorithm had the lowest prediction accuracy 5.990%, 6.85%, 2.332% are the MAPE percentages for the BTC, ETH, LTC respectively. In summary, the data indicates that the GRU algorithm yields the most accurate forecasts that correspond with the real cryptocurrency values, making it the best model for doing so. Despite the fact that using this approach is the only It is advised to invest at your own risk and do research to find cryptocurrency investors who appear to be seeing the greatest price gain for a certain coin.