Comparison of different methods used to forecast: An application on cryptocurrencies

In this study, using datasets of hourly and daily Bitcoin, Litecoin and Ethereum cryptocurrencies, forecasting values are obtained with the help of Integrated Autoregressive Moving Average (ARIMA), Artificial Neural Networks (ANN) and ATA methods, which have become increasingly important in time series analysis in recent years. results were compared. In determining the model that gives the closest result to the real values, the mean squares of error (MSE) values are taken into account, and the models are compared. In the analysis results, it was observed that the results obtained from the ATA, ANN and ARIMA models were the closest to the actual and the MSE values of these methods were the smallest, respectively. Detailed results obtained from the study are given in the relevant figures and tables. Keywords: Time series, Autoregressive integrated moving average, Artificial neural network, ATA method.