Effective Price Prediction of Cryptocurrencies using CNN-Based Dual Directional Model

Abstract

In recent days, predicting cryptocurrency trends has become essential for most individuals, including stockholders and traders, as it aids them in making better-informed selections regarding the digital asset market’s future. Predicting the price of the cryptocurrency leads to profitability in trading strategies. Therefore, the main objective of the study is to create an effective deep learning architecture using forecasting models such as recurrent neural networks (recurrent NN), Convolutional neural networks (convolutional NN) and Long Short-Term Memory (LSTM) for predicting Bitcoin and Ethereum prices. The model utilizes CNN to extract features from historical price data. Where, CNN has the ability to detect complex patterns in price movements and incorporates bidirectional LSTM (Bi-LSTM) layers to capture both past and future price trends. Moreover, Bi-LSTM effectively manages the temporal dynamics and makes it suitable for financial time series, which demonstrate non-linear behavior. Experimental results on a dataset of major cryptocurrencies demonstrate the efficacy of the proposed model in forecasting cryptocurrency prices with high accuracy. The dual-directional model outperforms traditional time series forecasting methods and singledirectional models, showcasing its potential for improving price prediction in the cryptocurrency stock market.