SVR OPTIMIZATION WITH PSO FOR CRYPTOCURRENCY PRICE PREDICTIONS

Abstract

Cryptocurrency is the nickname given to a system that uses Cryptography technology to securely transmit data and process digital currency exchanges in a dispersed manner. A Cryptocurrency is a form of risky investment, Cryptocurrency prices are very volatile (changing) making Cryptocurrency prices need to be predicted to make a profit. Support Vector Regression (SVR) is one method for predicting time series data such as Cryptocurrency prices. However, the SVR parameters need to be optimized to get accurate results. The Particle Swarm Optimization (PSO) algorithm is implemented to determine the effect on the optimization of SVR parameters. The implementation of SVR and SVR-PSO is carried out on Bitcoin and Shiba Inu Coin Cryptocurrency data. The result of this research is that the SVR algorithm has an accuracy of 13.19082% (Bitcoin) and 68.3221% (Shiba Inu Coin). The SVR-PSO algorithm obtained an accuracy of 96.92359% (BTC) and 94.74245% (SHIB).