Comparison of Forecasting Methods Using Time Series Models

Abstract

Forecasting needs to be done to predict conditions in the future so that they can prepare the required resources. PT XYZ is a manufacturer of veterinary drugs with one of its products being probiotics with a size of 2kg whole research problems being studied. Demand for probiotic products fluctuated so that forecasting was needed to predict the number of products in the future period. This study aims to compare demand forecasting with forecasting methods on the Time Series Forecasting Model and determine the best forecasting method. Forecasting methods used are Moving Average, Weighted Moving Average, Exponential Smoothing, and Trend Linear. Based on the error measurements that have been made, the Trend Linear method gives the Mean Square Error, Root Mean Square Error and Mean Percentage values smaller than other methods. Meanwhile, the error measurement based on the Mean Absolute Deviation and Mean Absolute Percentage Error, the Exponential Smoothing method with α = 0.5 gives a smaller value than other methods. However, in the verification of forecasting based on tracking signals, the results show that the Linear Trend method is more predictive and gives closer to actual results than the Exponential Smoothing method with α = 0.5. Therefore, the best forecasting method chosen is the Linear Trend method. The results of forecasting the demand for 2kg probiotic products for the next period show an increase in demand. The results of this study are expected to provide recommendations for companies to determine policies to prepare to fulfill demand in the coming period.