Parameter Optimization In Support Vector Regression Using Grid Search Optimization In Predicting Currency Values

Abstract

Currency Values is cryptocurrency which known has high price fluctuation. The more volatile the price of an investment, the higher the risk of investors gain losses. That principle also follows in Currency Values. In order for investments to avoid losses and gain profits, there needs a method that can be used to predict the price of Currency Values accurately. In this study, a Currency Values price prediction was carried out based on Currency Values data obtained in the past (time series forecasting) using the Support Vector Regression (SVR) method. The data used is weekly Currency Values price data from January 2018 to March 2020. Currency Values price data is nonlinear, so the Radial Basis Function kernel is used. The SVR parameter model is optimized using the Grid Search Optimization. The objective of this research is to get the accuracy of the SVR model by looking at the MAPE value which obtained. The results showed that the MAPE value obtained was 10.738% with the parameter values of C = 5, ? = 0.004, and ? = 0.07. The MAPE value indicates that the predicted results that obtained are categorized as good prediction results.