Comparison of different Artificial Neural Networks for Bitcoin price forecasting

-- This study investigates the impact of varying se - quen ce lengths on the accuracy of predicting cryptocurrency returns using Artificial Neural Networks (ANNs). Utilizing the Mean Absolute Error (MAE) as a threshold criterion, we aim to enhance prediction accuracy by excluding returns that are smaller than this threshold, thus mitigating errors associated with minor returns. The subsequent evaluation focuses on the accuracy of predicted returns that exceed this threshold. We compare four sequence lengths -- 168 hours (7 days), 72 hours (3 days), 24 hours, and 12 hours -- each with a return prediction interval of 2 hours. Our findings reveal the influence of sequence length on prediction accuracy and underscore the potential for optimized sequence configurations in financial forecasting models. Since Bitcoin was introduced in 2008 as a digital peer - to - peer equivalent currency built on blockchain technology [1], it emerged as a financial asset that is nowadays mainly used for investments [2]. The forecasting of time series data, such as the Bitcoin price, is a well - known problem existing in many different domains. Depending on the type of data being predicted, the difficulty of achieving an accurate result varies. For example, the prediction of the next sunrise time is relatively easy, whereas tomor - row's winning lottery numbers cannot be predicted with any accuracy. There are many methods for time series forecasting, ranging from classical mathematical models to approaches using d eep neural networks and deep learning [3]. In this paper, the data - driven approach of forecasting by applying different types of Artificial Neural Network (ANN) is used. We try to predict the future price movement of Bitcoin just by reviewing the past mark et data and compare the performance of the different ANNs on this task based on their predictions.