Deep Neural Networks with Noise Induction Effect for CryptocurrencyPrice Prediction

Over the past decade, with the advent of blockchain technology, we are witnessing a dramatic increase in the use of cryptocurrencies. However, investing in the cryptocurrency market is risky due to the market’s erratic behavior and high price volatility. Accordingly, the need to use an appropriate model for forecasting in risk control and management is considered intelligent. Motivated by the aforementioned issue, we propose a new approach based on a deep neural network focusing on the pattern of errors. The proposed approach is based on the random walk theory that argues that price movements are not all that random and that predictable component does indeed exist. This new approach, tries to improve the forecast results by modeling the residual values and inducing their effect on the main forecasts. We used the Long Short-Term Memory (LSTM) as the main prediction model and Vector Auto Regression (VAR) for noise prediction on three famous cryptocurrencies: Bitcoin, Ethereum, and BNB. The results show that the proposed approach has been able to improve forecasts.