Evaluation of bitcoin price changes before and after COVID-19 by machine learning, time series analysis and deep learning algorithms

Blockchain technology, which has been become quite widespread in use recently, has become very popular with the Internet technology. Bitcoin, which has been developed with blockchain technology, is the virtual currency that holds the most market volume among virtual currencies. Due to the lack of a central authority that controls the virtual currency markets, this market is open to price manipulations and external interventions, so that guidance is needed for the end-investor to invest. Recently, a number of methods have started to be been used to meet this need. In this study, various forecasting and classification methods about fluctuation in Bitcoin prices were evaluated together using machine learning, time-series analysis and deep learning methods. In this context, two separate datasets have been created based on the Bitcoin closing prices and up-to-down trends before and after the coronavirus pandemic. The success of forecasting and classification methods on these two datasets were evaluated and compared. As a result of the comparisons, Support Vector Machines method for the study conducted with the data before the pandemic, and ARIMA method for the study conducted with the data after the pandemic, had the most successful results.