Evolutionary strategy for learning multiple linear regression model in time series forecasting

• Main Authors: Susanto, Ardi, Afiahayati, Afiahayati, Abidin, Taufiq, Auzikri, Alvin
• Format: Conference Proceeding
• Language: English
• Subjects: Data analysis; Error analysis; Evolutionary algorithms; Forecasting; Regression analysis; Regression models
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0198860

This study aims to obtain the best model from data on the value of 4 currency pairs, namely USD/JPY, USD/CHF, GBP/USD and EUR/USD for the period January to December 2015 and data on foreign tourist visits to the DIY province for the period January 2006 to December. 2016. The data analysis technique in this research uses simple linear regression. By using the Evolutionary Strategy, the optimal value of the coefficient parameter is obtained in forecasting Multiple Linear Regression. In this study, Mean Square Error (MSE) and Mean Absolute Percentage Deviation (MAPD) were used to measure forecasting errors. The dataset used in this study is data on the value of 4 currency pairs, namely USD/JPY, USD/CHF, GBP/USD, and EUR/USD for the period January to December 2015 and data on foreign tourists visiting the province from January 2006 to December 2016. Based on the research results, the Evolutionary Strategy algorithm as an alternative method of finding coefficient values in Multiple Linear Regression produces a model with a small error value. The MAPD value obtained from testing with USD/JPY currency data is 0.5942%, USD/CHF is 0.4946%, GBP/USD is 0.5073% and EUR/USD is 0.4059%. Meanwhile, for foreign tourists, the MAPD value is 0.134%.