Prediction of Cryptocurrency States With Artificial Neural Networks and Support Vector Machines
Abstract
Tezde; Bitcoin (BTC), Ethereum (ETH), Binance Coin (BNB) ve Ripple (XRP) kripto paralarına ait 50 günlük ileriye dönük fiyat ve hacim tahminleri, makine öğrenmesinin Yapay Sinir Ağları (ANN) ve Destek Vektör Makineleri (SVM) yöntemleri kullanılarak incelenmiştir. Hangi yönteminin daha güçlü sonuçlar verdiği Hata Kareler Ortalamasının Karekökü (RMSE), Ortalama Mutlak Hata (MAE) ve 1 Geçikmeli Hataların Otokorelasyonu (ACF1) bilgi kriterleri kullanılarak belirlenmiştir. Analiz sonuçları; ANN modelinin BTC, BNB, ETH ve XRP için fiyat ve hacim tahmininde daha iyi sonuçlar verdiğini göstermiştir. 50 günlük fiyat tahmin sonuçlarına göre; BTC için ANN modelinde düşüşün SVM modelinde ise bir artışın olduğu, BNB için ANN modelinde fiyatın bir pik yapmasının ardından düşüşün meydana geldiği, SVM modelinde ise bu artışın daha istikrarlı olduğu, ETH için ANN modelinde fiyatın yatay bir harekete sahip olduğu ve stabil hareketlenmelerin meydana geldiği, SVM modelinde ise düşüşün meydana geldiği ve ETH fiyat tahmininde ANN ve SVM modelinden elde edilen sonuçların XRP'de de aynı sonuçlar verdiği gözlemlenmiştir. 50 günlük hacim tahmin sonuçlarına göre ise; BTC için ANN ve SVM modellerinde artışın meydana geldiği, BNB için ANN modelinde düşüşün meydana geldiği, SVM modelinde ise bir artışın varlığı, ETH için ANN modelinde hacmin yatay bir harekete sahip olduğu ve stabil hareketlenmelerin meydana geldiği gözlemlenirken SVM modelinde ise hızlı bir artışın meydana geldiği ve son olarak XRP hacim tahmininde ANN modelinde yatay bir harekete sahip olduğu ve stabil hareketlenmelerin meydana geldiği gözlemlenirken SVM modelinde bir düşüşün meydana geldiği görülmüştür.
In the thesis; the 50-day forward price and volume forecasts for Bitcoin (BTC), Ethereum (ETH), Binance Coin (BNB) and Ripple (XRP) coins were analyzed using Artificial Neural Networks (ANN) and Machine Support Vectors (SVM) methods of machine learning. Which method gave stronger results was determined by using Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Autocorrelation of errors at lag 1 (ACF1) information criteria. Analyzes were made using the R program and data on cryptocurrencies are obtained from the web address investing.com. As a result of the analysis, it has been observed that the ANN model gives better results for BTC, BNB, ETH and XRP in forward price and volume prediction. According to the 50-day price estimation results obtained; While there is a decrease in the ANN model for BTC, there is an increase in the SVM model, a decrease occurs after the price peaks in the ANN model for BNB, this increase is more stable in the SVM model, the price has a horizontal movement in the ANN model for ETH and stable movements occur. It has been observed that there is a decrease in the SVM model, and the results obtained from the ANN and SVM model in ETH price prediction give the same results in XRP. According to the 50-day volume estimation results; it is observed that an increase occurred in the ANN and SVM models for BTC, a decrease occurred in the ANN model for BNB, an increase in the SVM model, a horizontal movement in the volume in the ANN model for ETH and stable movements, while a rapid increase occurred in the SVM model and Finally, in XRP volume estimation, it was observed that there was a horizontal movement in the ANN model and stable movements were observed, while a decrease occurred in the SVM model.
In the thesis; the 50-day forward price and volume forecasts for Bitcoin (BTC), Ethereum (ETH), Binance Coin (BNB) and Ripple (XRP) coins were analyzed using Artificial Neural Networks (ANN) and Machine Support Vectors (SVM) methods of machine learning. Which method gave stronger results was determined by using Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Autocorrelation of errors at lag 1 (ACF1) information criteria. Analyzes were made using the R program and data on cryptocurrencies are obtained from the web address investing.com. As a result of the analysis, it has been observed that the ANN model gives better results for BTC, BNB, ETH and XRP in forward price and volume prediction. According to the 50-day price estimation results obtained; While there is a decrease in the ANN model for BTC, there is an increase in the SVM model, a decrease occurs after the price peaks in the ANN model for BNB, this increase is more stable in the SVM model, the price has a horizontal movement in the ANN model for ETH and stable movements occur. It has been observed that there is a decrease in the SVM model, and the results obtained from the ANN and SVM model in ETH price prediction give the same results in XRP. According to the 50-day volume estimation results; it is observed that an increase occurred in the ANN and SVM models for BTC, a decrease occurred in the ANN model for BNB, an increase in the SVM model, a horizontal movement in the volume in the ANN model for ETH and stable movements, while a rapid increase occurred in the SVM model and Finally, in XRP volume estimation, it was observed that there was a horizontal movement in the ANN model and stable movements were observed, while a decrease occurred in the SVM model.
Description
Keywords
İstatistik, Destek vektör makineleri, Ethereum, Kripto para birimi, Ripple, Yapay sinir ağları, Statistics, Support vector machines, Ethereum, Crypto currency, Ripple, Artificial neural networks
Turkish CoHE Thesis Center URL
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
92

