| dc.description |
Cryptocurrencies, which entered our lives in the recent past and found a place in the financial markets in a short time, are used both as a means of exchange and an investment tool. The fact that cryptocurrencies are not under the control of a central authority has brought about fluctuations in the prices of these tools. Therefore, the development of an intelligent forecasting model is very important for the selection of financial assets to be invested and the realization of investment decisions. Deep learning and artificial intelligence are used in the selection of cryptocurrency and other investment instruments to be invested. Deep learning models such as the Recurrent Neural Network (RNN), Long-Short Term Memory (LSTM) and the Gated Recurrent Unit (GRU) model have been proven by researchers to outperform traditional time series models in cryptocurrency price prediction. For this reason, in this study, a 30-day price estimate of Bitcoin, Ethereum and Ripple, which are the cryptocurrencies with the highest market value and transaction volume, has been made using LSTM and GRU, a special RNN method. As a result of the research, Bitcoin gave the best prediction result in both models. The second best prediction result was found for Ripple, then Ethereum. When the methods used were compared, the best estimation result was reached with the GRU model for Bitcoin and Ripple, and the LSTM model for Ethereum, according to the MAPE performance criterion. |
|