Time-Series Analysis of Cryptocurrency Price: Bitcoin as a Case Study

Abstract

Bitcoin (BTC) is a distributed virtual paradigm that uses a peer-to-peer network to provide a means of digital money. Bitcoin pricing was changing monthly, and the BTC price observations have been collected since October 2013. In this paper, we propose a neural network-based autoregressive predictive model to forecast the monthly pricing of cryptocurrency bitcoin technology based on 100 historical observations for the bitcoin prices from Oct-2013 to Sep-2021 (in us dollars). Specifically, the proposed scheme uses a nonlinear autoregressive neural network with external input (NARX) by detaining the maximum regression coefficient corresponding to the most prediction accuracy and the least normalized prediction error. The simulation results showed that the highest prediction accuracy for the identified cryptocurrency, bitcoin pricing is 99.1%. The subsequent perdition model was effectively used to anticipate the evolution of forthcoming 12-month data records for the n cryptocurrency bitcoin prices from Oct-2021 to Sep-2022. The forecast values reveal a very slow, linearly developing tendency in the prices of cryptocurrency bitcoin released monthly over the past ten years' records for the global cryptocurrency bitcoin pricing time series.