Forecasting volatility of traditional financial instruments is a well known and widely addressed problem. In the past, researches addressed it by using technical and fundamental analysis. The former looks at the past price movement of a currency or a stock (their market value and trading volumes), while the latter analyses outside information which can cause fluctuations in the currency or stock value (e.g.: introducing a new product in the company can increase the value of company’s stocks, while lower rating of a country can reduce the value of its currency). The present master’s thesis focuses on the latter, i.e. fundamental analysis. Its effectiveness will be demonstrated on a fairly new and not yet well established digital currency market. The main advantage introduced by the digital currency market — in comparison with traditional financial markets — is its P2P nature. This means that every user has an insight into the entire trading process (market orders, demands, offers, transactions). Moreover, the user has access to all the information regarding the functioning of the network (computer power consumption, amount of currency in circulation, numbers of miners...).
The first, introductory part of the thesis offers an overview of existing methods for predicting currency fluctuations that were adopted which mostly come from the field of trading with securities. Moreover, a rough presentation of our improvements and the advantages of the digital currency market are presented. Follows a detailed description on how the digital currency market functions: the process of transactions is described, the role of miners in the Bitcoin Network and the process of verification of users and transactions are explained, and the possibilities of trading with digital currencies are shown. In the next chapter the adopted development environment is described (how the application is built, tools that were used and libraries). The central part of the thesis demonstrates the development of our proposed method, the goal of which is to predict price movements of Bitcoin. This part is divided into three main parts: data mining, analysis of the considered data and the simulation. In the first part the web resources and methods of data collecting are defined. In the second part, an analysis of the data collected is conducted, on the basis of which only the data that could influence the value of currency is selected. Lastly, the selected data is implemented in a tool simulation which predicts currency fluctuations in which two models were applied: multiple linear regression and artificial neural network.