Predictive Analysis of the Cryptocurrencies’ Movement Direction Using Machine Learning Methods
Küçük Resim Yok
Tarih
2021
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Springer Science and Business Media Deutschland GmbH
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Cryptocurrencies are among the most interesting financial instruments of recent years. Unlike the classical understanding that money exists as a means of change from one hand to another, this digital economy has begun to attract people's attention. The most popular currency emerging from this concept of cryptocurrency is “Bitcoin”. As the popularity of Bitcoin started to increase since 2016, the number of academic studies on cryptocurrencies has increased in parallel. In light of these developments, our study proposes predictive models of price change directions of high market value cryptocurrencies. Bitcoin, Ethereum and Litecoin were selected as cryptocurrencies and daily opening, closing, high and low prices of these currencies were collected from financial websites. Preprocessing was performed on the collected data to create input vectors. These vectors were given regression algorithms which are Multiple Linear, Polynomial, Support Vector, Decision Tree and Random Forest Regression. As evaluation metrics, R-square Method (R2) and Root Mean Square Error (RMSE) were chosen. After doing experiments with different parameter settings, it was found out that the chosen machine learning models showed satisfactory performances in predicting the directions of then mentioned cryptocurrencies. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Açıklama
Anahtar Kelimeler
Cryptocurrency, Data mining, Prediction, Regression
Kaynak
Lecture Notes on Data Engineering and Communications Technologies
WoS Q Değeri
Scopus Q Değeri
Q3
Cilt
76