COMPARISON OF RANDOM FOREST AND SINGLE EXPONENTIAL SMOOTHING METHODS IN THE PREDICTION OF GOLD DEMAND FOR INDONESIAN JEWELRY PERIOD 2010-2021

Abstract

Forecasting is an activity to predict something that will happen in the future. In forecasting, there are several methods, including Random Forest and Single Exponential Smoothing (SES). Random Forest has the advantage that it does not require assumptions, while SES requires assumptions as data that must be stationary and used for short-term forecasting. We can apply Random Forest and SES in various fields, one of which is economics, which focuses on the demand for gold jewellery. The importance of analysing data on the number of requests for gold jewellery is because there are many enthusiasts from gold itself. This study compares the two methods in predicting the demand for gold jewellery in Indonesia. Based on the results of the analysis, we found Random Forest is better than SES, we can see it from the forecast error value of Random Forest, which is smaller than SES. The Random Forest method got the best model with n-tree 50, resulting in an MAPE value of 12.87%. Meanwhile, the best model for SES with an alpha of 0.4 produces an MAPE value of 20.63%.