| Abstract [eng] |
Observing the economic changes in recent years worldwide and the unprecedented factors related to the pandemic and the war in Ukraine, investors are increasingly seeking ways to avoid the problems of inflation and currency depreciation. While stock markets, commodities, securities, currencies, and other investment products continue to attract stable investor attention, a relatively recent market, cryptocurrencies, is rapidly gaining popularity, generating increased interest as an investment. This study aims to explore the relationships between external factors and changes in the price of Bitcoin. The objective of this work is to conduct a study on the impact of external factors on the analysis of Bitcoin prices (investment returns), which would help summarize investor behavior. Based on the economic literature analysis conducted in the first part of this work, which summarizes the differences between the financial product market and the consumer goods market, it has been determined that consumer behavior has undergone significant changes following the COVID-19 pandemic, with new habits in consumption, saving, and investing being observed. Furthermore, the focus of the study is to examine the changes in the cryptocurrency market observed from 2013 to 2023. It can be concluded that since 2013, the cryptocurrency market has changed beyond recognition, witnessing the growth of major cryptocurrency capitalizations, with Bitcoin still occupying the largest market share. The volatility of Bitcoin prices and its sudden jumps and drops are difficult to calculate and predict for the future. Additionally, this study includes an analysis of speculative activities by consumers, involving prearranged agreements and illegal activities related to cryptocurrency investments. The second part of the work presents descriptions of mathematical models applied in the study, based on mathematical literature. Finally, in the third part of the work, an analysis of Bitcoin prices and the prices of external factors is performed. Investment product returns are calculated, a stylized fact analysis of the Bitcoin time series is conducted, GARCH modeling is performed, correlation analysis is carried out, and machine learning methods are employed to address the regression problem. During GARCH modeling, a model is created, and GARCH residuals of Bitcoin prices are calculated. Furthermore, the correlation analysis indicates that there are no easily identifiable linear relationships between the modeled Bitcoin returns by GARCH models and the logarithmic returns of external factors. This relationship is weak, nonexistent, or nonlinear. Ultimately, mathematical machine learning methods are employed using Random Forest, XGBoost, and Neural Network algorithms. It is found that the \textit{XGBoost} algorithm performs the best in predicting the regression problem of Bitcoin returns, with the most significant variables being the lagged Bitcoin returns and the lagged London Robusta coffee and Apple company. |
