Bitcoin Fiyat Değişimlerinin Makine Öğrenmesi Yöntemi ile Tahmin Edilmesi
Prediction of the Bitcoin Price Changes Through Machine Learning

Summary/Abstract: Purpose – The rapid changes in the prices of cryptoassets, especially Bitcoin, attract attention by both financial investors and the media. Accordingly, many researchers and financial actors, with several different motivations, especially with the aim of making a profit, are trying to determine various factors that affect the price of Bitcoin. A detailed examination is carried out on attributes such as the Fed Interest rate, gold and Bitcoin's different price indicators, which are thought to affect Bitcoin price movements. In this context, a systematic analysis is conducted on various machine learning algorithms used to predict prices. Design/Methodology/Approach – Four models were used, different estimation error rates were obtained, and it was seen that each of them could be used in the study. Findings – The analysis results show that the best prediction performance recommended for the Bitcoin dataset is as follows, respectively: random tree (RF) 96.38%, decision tree (DT) 96.28%, linear regression 95.06 and Stochastic Gradient Descent (SGD) linear regression 93.91%. It has been concluded that Bitcoin price changes are more highly affected by their own price changes, rather than the Fed interest rate and gold. Discussion – According to the two algorithms that give the best results in machine learning algorithms.İt can be said that th intraday high price is extremely effective. The lowest price is the second most affective attribute. The result in question shows that Bitcoin is most affected by its own price fluctations.

• Details
• Contents