Modelling and Forecasting the Trend in Cryptocurrency Prices

Abstract

The prediction of cryptocurrency prices is a hot topic among academics. Nevertheless, predicting the cryptocurrency price accurately can be challenging in the real world. Numerous studies have been undertaken to determine the best model for successful prediction. However, they lacked correct results because they avoided identifying the critical features. It is important to remember that trends are critical features in time series to obtain data information. A dearth of research demonstrates that the cryptocurrency trend comprises linear and nonlinear patterns. Therefore, this study attempted to fill this gap and focused on modelling and forecasting trends in cryptocurrency. This study examined the linear and nonlinear dependency trend patterns of the top five cryptocurrency closing prices. The weekly historical data of each cryptocurrency were taken at different periods due to the availability of data on the system. In achieving its goal, this study examined the results by plotting based on residual trend and diagnostic statistic checking using three deterministic methods: linear trend regression, quadratic trend, and exponential trend. Based on the minimum Akaike Information Criterion (AIC), the result showed that the top five cryptocurrency closing price data series contained nonlinear and linear trend patterns. The information of this study will assist traders and investors in comprehending the trend of the top five cryptocurrencies and choosing the suitable model to predict cryptocurrency prices. Additionally, accurately measuring the forecast will protect investors from losing their investment.