ANALYSIS AND COMPARING FORECASTING RESULTS USING TIME SERIES METHOD TO PREDICT SALES DEMAND ON COVID-19 PANDEMIC ERA

Abstract

The Covid-19 pandemic has made uncertainty in demand very high; there have been many changes in demand due to changes in the market and people's buying methods. So that forecasting accuracy is significant for every industry, at least the forecast that is closest to the conditions faced by the company so that the company does not lose money due to forecasting errors. Time series is a widely used model for forecasting using past data. This study aims to minimize forecasting errors by analyzing which demand forecasting model is most suitable for demand conditions based on historical data on demand for masterbatch products. The method used in this study is a time series model, which consists of the season naive method, holt exponential smoothing, exponential triple smoothing, and autoregressive integrated moving average (ARIMA). Data processing is done using Rstudio software. The results show that the ARIMA method (2,1,0) (1,1,0) is the best because it has the smallest error rate value with case studies and exact data; the standard error size values used are ME, RMSE, MAE, MPE, MAPE, MASE, and ACF1. This study analyzes forecasting during the Covid-19 pandemic using time series and compares them to find the best results. Then the results of this study can be used as a reference by companies and researchers in determining the model used to make forecasts.