Application of Multivariate Singular Spectrum Analysis for Forecasting the Production of Plantation Commodities

Abstract

This study aims to determine the production forecast of plantation commodities in North Sumatera Province, including palm oil, rubber, coffee, cocoa, and tobacco. Therefore, a forecasting method that can capture patterns and interrelationships between variables simultaneously is needed. This study applies the Multivariate Singular Spectrum Analysis (MSSA) method in forecasting the production of several major plantation commodities in North Sumatra Province. The results obtained from the MSSA analysis stage are a window length (l) of 2 and r-grouping of 2 with a forecasting period length of 2 periods in chronological order. The forecasting results based on the forecasting accuracy level using MAPE for the production of palm oil, rubber, coffee, cocoa, and tobacco in North Sumatra Province using the Multivariate Singular Spectrum Analysis method are 3.57223%, 3.95038%, 6.92317%, 3.03589%, and 3.03589%. Based on the MAPE accuracy category for each variable, the MAPE values are < 10% and fall into the accurate category for forecasting.