<![CDATA[Refillable drinking water depots often face challenges in dealing with unpredictable customer demand on a daily basis. This uncertainty complicates the process of stock management, production planning, and overall operations. Without accurate sales forecasts, depots risk losing potential sales and experiencing a decline in service quality to customers. Therefore, a solution is needed that can accurately predict daily sales. The first step in this research is to collect relevant data. Once the data is available, pre-processing is conducted to prepare the data before entering the modeling process. The Long Short-Term Memory (LSTM) model has the advantage of remembering historical patterns in time series data. Meanwhile, the Seasonal Autoregressive Integrated Moving Average (SARIMA) model is an extension of ARIMA that can handle data with seasonal characteristics. In this study, the LSTM model demonstrated better performance than SARIMA. This is evidenced by the performance evaluation values: MAPE of 9.54%, RMSE of 0.17, and MAE of 0.14 for the LSTM model, which are lower than MAPE of 10.51%, RMSE of 0.19, and MAE of 0.16 for SARIMA. These values indicate that LSTM is capable of providing more accurate prediction results. Based on these results, it can be concluded that the LSTM model is more effective and recommended for use in predicting daily sales of refillable water at the Manshurin Water depot]]>
Copyrights © 2025
