Sales Forecasting for Fresh Foods: A study in Indonesian FMCG

摘要

This study aims to get the accurate bread demand forecasting method in one of the major bread producers in Indonesia. Given the many types of bread sold, this study is limited to five types of products. Using several time-series forecasting methods used, including the moving average method, the exponential smoothing method, multiple regression methods, and the SVR method (Support Vector Regression). Forecasting bread sales using the best method, which is the method that produces the smallest error value. Forecasting error methods used are the Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and Mean Squared Error (MSE). The results show that to produce high forecasting accuracy, the forecasting method must be supported by the data mining process. This research empirically proves that store classification combined with seasonal factors such as weekends, holidays, and pay periods have an influence on forecasting accuracy.