Jurnal - Komparasi Kinerja Holt-Winters dan Simple Moving Average pada Data Penjualan Cetakan Berbasis Pra-Pemrosesan Winsorize

Abstract

Sales forecasting plays an important role in production planning and inventory management in the printing industry. This study aims to compare the performance of the Holt-Winters (Triple Exponential Smoothing) and Simple Moving Average (SMA) methods in predicting sales based on real printing transaction data from August 2022 to November 2023. The data were aggregated weekly and underwent preprocessing using Winsorizing and Epsilon Adjustment to handle extreme value anomalies and Business-to-Business (B2B) order fluctuations. Subsequently, the data were divided into 80% training data and 20% testing data. Model evaluation was conducted using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results indicate that the automatically optimized Holt-Winters method outperforms the SMA (Window 3) method, yielding a lower error rate with a MAPE value of 48.84% compared to SMA's 54.30%. This finding proves that with proper preprocessing, printing sales data exhibit a 4-week recurring seasonal pattern successfully captured by the Holt-Winters model. The novelty of this study lies in the application of Winsorizing techniques to highly volatile real-world printing industry data to stabilize variance, along with a comparative analysis to determine the most adaptive forecasting model for the supply chain